diff --git a/.claude/skills/address-pr-comments/SKILL.md b/.claude/skills/address-pr-comments/SKILL.md
index 9a86f80b..d31e1ced 100644
--- a/.claude/skills/address-pr-comments/SKILL.md
+++ b/.claude/skills/address-pr-comments/SKILL.md
@@ -16,18 +16,32 @@ Determine the target PR:
 
 ```bash
 # If argument provided, use it; otherwise detect from current branch
-gh pr view $ARGUMENTS --json number,url,headRefName,baseRefName
+gh pr view $ARGUMENTS --json number,url,headRefName,baseRefName,author
 ```
 
 If no PR is found, stop and inform the user.
 
-Extract owner, repo, and PR number for subsequent API calls:
+Extract owner, repo, PR number, and **PR author login** for subsequent API calls:
 
 ```bash
 gh repo view --json owner,name --jq '"\(.owner.login)/\(.name)"'
 ```
 
-### 2. Fetch all review comments
+### 2. Fetch review comments and summaries
+
+#### 2a. Determine the latest review round
+
+Find the timestamp of the most recent push to the PR branch — this marks the boundary of the current review round:
+
+```bash
+# Get the most recent push event (last commit pushed)
+gh api repos/{owner}/{repo}/pulls/{pr-number}/commits \
+  --jq '.[-1].commit.committer.date'
+```
+
+Store this as `$LAST_PUSH_DATE`. Comments created **after** this timestamp are from the current (latest) review round. If no filtering by round is desired (e.g., first review), process all unresolved comments.
+
+#### 2b. Fetch inline review comments
 
 Retrieve all review comments (inline code comments) on the PR:
 
@@ -38,18 +52,28 @@ gh api repos/{owner}/{repo}/pulls/{pr-number}/comments \
   2>&1 | head -500
 ```
 
-Also fetch top-level review comments (review bodies):
+#### 2c. Fetch review summaries
+
+Fetch top-level review comments (review bodies/summaries). These often contain high-level feedback and action items:
 
 ```bash
 gh api repos/{owner}/{repo}/pulls/{pr-number}/reviews \
-  --jq '.[] | select(.body != "" and .body != null) | {id: .id, user: .user.login, state: .state, body: .body}' \
+  --jq '.[] | select(.body != "" and .body != null) | {id: .id, user: .user.login, state: .state, body: .body, submitted_at: .submitted_at}' \
   2>&1 | head -200
 ```
 
-Filter out:
-- Comments authored by the PR author (self-comments, unless they contain a TODO/action item)
+**Pay special attention to review summaries** — they often list multiple action items in a single review body. Parse each action item from the summary as a separate work item.
+
+#### 2d. Filter comments
+
+**Include** comments from:
+- **Reviewers** (anyone who is not the PR author) — standard review feedback
+- **The PR author themselves** — self-comments are treated as actionable TODOs/notes-to-self that should be addressed
+- **@codex** and other AI reviewers — treat their comments with the same weight as human reviewer comments
+
+**Exclude**:
 - Already-resolved threads
-- Bot comments that are purely informational
+- Bot comments that are purely informational (CI status, auto-generated labels, etc.) — but NOT @codex or other AI reviewer comments, which are substantive
 
 Check which threads are already resolved:
 
@@ -79,6 +103,28 @@ gh api graphql -f query='
 
 Only process **unresolved** threads with actionable comments.
 
+#### 2e. Prioritize latest comments
+
+When there are many unresolved comments, prioritize:
+1. Comments from the **latest review round** (after `$LAST_PUSH_DATE`)
+2. Comments from review summaries (they represent the reviewer's consolidated view)
+3. Older unresolved comments that are still relevant
+
+### 2b. Read the PR specs
+
+Before evaluating any comment, read the PR description to check for a **SPECS** section:
+
+```bash
+gh pr view $ARGUMENTS --json body --jq '.body'
+```
+
+If a SPECS section is present, **it defines the authoritative requirements for this PR**. Specs override:
+- Your assumptions about backward compatibility or design intent
+- Inline code comments
+- Conventions from other parts of the codebase
+
+Store the specs for use in step 4 (validity evaluation). If a reviewer comment aligns with a spec, the comment is **valid by definition** — even if you think the current implementation is reasonable.
+
 ### 3. Understand each comment
 
 For each unresolved review comment:
@@ -99,36 +145,45 @@ For each unresolved review comment:
 | **Nitpick** | Minor optional suggestion | Evaluate — fix if trivial, otherwise reply explaining the tradeoff |
 | **Invalid/outdated** | Comment doesn't apply or is based on a misunderstanding | Reply politely explaining why |
 
-### 4. Evaluate validity — bash behavior is the source of truth
+### 4. Evaluate validity — specs and bash behavior are the sources of truth
+
+There are two sources of truth, checked in this order:
 
-**The shell must match bash behavior unless it intentionally diverges** (e.g., sandbox restrictions, blocked commands, readonly enforcement). This principle overrides reviewer suggestions.
+1. **PR specs** (from step 2b) — if present, specs are the highest authority for what this PR should do
+2. **Bash behavior** — the shell must match bash unless it intentionally diverges (sandbox restrictions, blocked commands, readonly enforcement)
+
+**CRITICAL: Never invent justifications for dismissing a comment.** If a reviewer says "the spec requires X" and the spec does require X, the comment is valid — even if you think the current implementation is a reasonable alternative. Do not fabricate reasons like "backward compatibility" or "design intent" unless those reasons are explicitly stated in the specs or CLAUDE.md.
 
 For each comment, determine if it is **valid and actionable**:
 
-1. **Verify against bash** — always check what bash actually does:
+1. **Check against PR specs first** — if a SPECS section exists and the comment aligns with a spec, the comment is **valid by definition**. Do not dismiss it.
+2. **Verify against bash** — for comments about shell behavior, check what bash actually does:
    ```bash
    docker run --rm debian:bookworm-slim bash -c '<relevant script>'
    ```
-2. **Read the relevant code** in full — not just the diff, but the surrounding implementation
-3. **Check project conventions** in `CLAUDE.md` and `AGENTS.md`
-4. **Consider side effects** — will the change break other tests or behaviors?
-5. **Check for duplicates** — is the same issue raised in multiple comments? Group them
+3. **Read the relevant code** in full — not just the diff, but the surrounding implementation
+4. **Check project conventions** in `CLAUDE.md` and `AGENTS.md`
+5. **Consider side effects** — will the change break other tests or behaviors?
+6. **Check for duplicates** — is the same issue raised in multiple comments? Group them
 
 Decision matrix:
 
-| Reviewer says | Bash does | Shell intentionally diverges? | Action |
-|--------------|-----------|-------------------------------|--------|
-| "This is wrong" | Reviewer is right | No | **Fix the implementation** to match bash |
-| "This is wrong" | Current code matches bash | No | **Reply** explaining it matches bash, with proof |
-| "This is wrong" | N/A | Yes (sandbox/security) | **Reply** explaining the intentional divergence |
-| "Do it differently" | Suggestion matches bash better | No | **Fix the implementation** to match bash |
-| "Do it differently" | Current code already matches bash | No | **Reply** — bash compatibility takes priority |
+| Reviewer says | Spec says | Bash does | Action |
+|--------------|-----------|-----------|--------|
+| "Spec requires X" | Spec does require X | N/A | **Fix the implementation** to match the spec |
+| "Spec requires X" | No such spec exists | N/A | **Reply** noting the spec doesn't mention this |
+| "This is wrong" | No spec relevant | Reviewer is right | **Fix the implementation** to match bash |
+| "This is wrong" | No spec relevant | Current code matches bash | **Reply** explaining it matches bash, with proof |
+| "This is wrong" | No spec relevant | N/A (sandbox/security) | **Reply** explaining the intentional divergence |
+| "Do it differently" | No spec relevant | Suggestion matches bash better | **Fix the implementation** to match bash |
+| "Do it differently" | No spec relevant | Current code already matches bash | **Reply** — bash compatibility takes priority |
 
 If a comment is **not valid**:
 - Prepare a polite reply with proof (e.g., "This matches bash behavior — verified with `docker run --rm debian:bookworm-slim bash -c '...'`")
 - If the divergence is intentional, explain why (sandbox restriction, security, etc.)
+- **Never claim "backward compatibility" or "design intent" unless you can point to a specific line in the specs or CLAUDE.md that says so**
 
-If a comment is **valid** (i.e., fixing it brings the shell closer to bash, or addresses a real bug):
+If a comment is **valid** (i.e., it aligns with a spec, brings the shell closer to bash, or addresses a real bug):
 - Proceed to step 5
 
 ### 5. Implement fixes
@@ -181,7 +236,11 @@ If fixes span unrelated areas, prefer multiple focused commits over one large co
 
 **All replies MUST be prefixed with `[<LLM model name>]`** (e.g. `[Claude Opus 4.6]`) so reviewers can tell the response came from an AI.
 
-For each comment that was addressed:
+Handle comments differently based on who authored them:
+
+#### Reviewer comments (not the PR author)
+
+For each reviewer comment that was addressed:
 
 1. **Reply** explaining what was fixed:
    ```bash
@@ -221,16 +280,57 @@ For each comment that was addressed:
    ' -f threadId="<thread-id>"
    ```
 
-For comments that were **not valid** or were **questions**, reply (prefixed with `[<MODELL NAME - VERSION>]`) with an explanation but do NOT resolve — let the reviewer decide.
+#### PR author self-comments
+
+For comments authored by the PR author (self-notes/TODOs):
+
+1. **Fix the issue** described in the comment (these are actionable items the author left for themselves)
+2. **Resolve** the thread (the PR author can resolve their own threads)
+3. **Do NOT reply** to self-comments — just fix and resolve. No need for the AI to narrate back to the same person who wrote the note.
+
+#### Review summary action items
+
+For action items extracted from review summaries (step 2c):
+
+1. **Fix each action item** as if it were an inline comment
+2. **Reply to the review** with a summary of all action items addressed:
+   ```bash
+   gh api repos/{owner}/{repo}/pulls/{pr-number}/reviews/{review-id}/comments \
+     -f body="[<MODEL NAME> - <VERSION>] Addressed the following from this review:
+   - <action item 1>: <what was done>
+   - <action item 2>: <what was done>"
+   ```
+   If the `comments` endpoint doesn't work for review-level replies, use an issue comment instead:
+   ```bash
+   gh api repos/{owner}/{repo}/issues/{pr-number}/comments \
+     -f body="[<MODEL NAME> - <VERSION>] Addressed review feedback from @{reviewer}:
+   - <action item 1>: <what was done>
+   - <action item 2>: <what was done>"
+   ```
+
+#### Invalid or question comments
+
+For comments that were **not valid** or were **questions**, reply (prefixed with `[<MODEL NAME> - <VERSION>]`) with an explanation but do NOT resolve — let the reviewer decide.
+
+**IMPORTANT: Never resolve a thread where the reviewer's comment aligns with a PR spec but the implementation doesn't match.** These are valid spec violations — fix the code instead. If you cannot fix it, leave the thread unresolved and explain the blocker.
 
 ### 8. Summary
 
-Provide a final summary:
+Provide a final summary organized by source:
+
+**Reviewer inline comments addressed:**
+- List each comment with: the comment (abbreviated), classification (bug, style, suggestion, etc.), what was changed
+
+**Review summary action items addressed:**
+- List each action item from review summaries that was implemented
+
+**PR author self-comments addressed:**
+- List each self-note/TODO that was fixed and resolved
+
+**Not fixed (with reason):**
+- List any comments replied to but not fixed, with explanation
+
+**Could not be addressed:**
+- List any comments that could not be addressed, with explanation
 
-- List each review comment that was addressed with:
-  - The comment (abbreviated)
-  - The classification (bug, style, suggestion, etc.)
-  - What was changed
-- List any comments that were replied to but not fixed (with reason)
-- List any comments that could not be addressed (with explanation)
-- Confirm the commit(s) pushed and threads resolved
+Confirm the commit(s) pushed and threads resolved.
diff --git a/.claude/skills/code-review/SKILL.md b/.claude/skills/code-review/SKILL.md
index 84d52ae5..d24f6778 100644
--- a/.claude/skills/code-review/SKILL.md
+++ b/.claude/skills/code-review/SKILL.md
@@ -26,7 +26,36 @@ git diff main...HEAD
 
 If no changes are found, inform the user and stop.
 
-### 2. Read and understand all changed code
+### 2. Verify specs implementation
+
+Read the PR description and look for a **SPECS** section:
+
+```bash
+gh pr view $ARGUMENTS --json body --jq '.body'
+```
+
+If a SPECS section is present, it defines the requirements that this PR MUST implement. **Every single spec must be verified against the diff.**
+The specs override other instructions (code, inline comments in code, etc). ALL specs MUST be implemented.
+
+For each spec:
+1. **Find the code** that implements the spec
+2. **Verify correctness** — does the implementation fully satisfy the spec?
+3. **Check for missing specs** — is any spec not implemented at all?
+
+Flag any unimplemented or partially implemented spec as a **P1 finding** (missing functionality that was explicitly required).
+
+Include a spec coverage table in the review output:
+
+```markdown
+| Spec | Implemented | Location | Notes |
+|------|:-----------:|----------|-------|
+| Must support `--flag` option | Yes | `interp/api.go:42` | Fully implemented |
+| Must return exit code 2 on error | **No** | — | Not found in diff |
+```
+
+If no SPECS section is found in the PR description, skip this step.
+
+### 3. Read and understand all changed code
 
 For each changed file:
 
diff --git a/.claude/skills/review-fix-loop/SKILL.md b/.claude/skills/review-fix-loop/SKILL.md
index 6a403dfb..d9f1bbe5 100644
--- a/.claude/skills/review-fix-loop/SKILL.md
+++ b/.claude/skills/review-fix-loop/SKILL.md
@@ -17,7 +17,7 @@ You MUST follow this execution protocol. Skipping steps or running them out of o
 Your very first action — before reading ANY files, before running ANY commands — is to call TaskCreate exactly 11 times, once for each step/sub-step below. Use these exact subjects:
 
 1. "Step 1: Identify the PR"
-2. "Step 2: Run the review-fix loop"
+2. "Step 2: Run the review-fix loop" ← **Update subject with iteration number each loop** (e.g. "Step 2: Run the review-fix loop (iteration 1)")
 3. "Step 2A1: Self-review (code-review)" ← **parallel with 2A2**
 4. "Step 2A2: Request external reviews (@codex)" ← **parallel with 2A1**
 5. "Step 2B: Address PR comments (address-pr-comments)"
@@ -91,7 +91,9 @@ Store the owner and repo name.
 
 **GATE CHECK**: Call TaskList. Step 1 must be `completed`. Set Step 2 to `in_progress`.
 
-Set `iteration = 1`. Maximum iterations: **10**. Repeat sub-steps A through E while `iteration <= 10`:
+Set `iteration = 1`. Maximum iterations: **30**. Repeat sub-steps A through E while `iteration <= 30`.
+
+**At the start of each iteration**, update the Step 2 task subject to include the current iteration number using TaskUpdate, e.g. `"Step 2: Run the review-fix loop (iteration 3)"`.
 
 ---
 
@@ -107,7 +109,11 @@ This analyzes the full diff against main, posts findings as a GitHub PR review w
 
 Post a comment to trigger @codex reviews:
 ```bash
-gh pr comment <pr-number> --body "@codex review"
+gh pr comment <pr-number> --body "@codex review this PR
+
+Important: Read the SPECS section of the PR description. If SPECS are present: **make sure the implementation matches ALL the specs**.
+The specs override other instructions (code, inline comments in code, etc). ALL specs MUST be implemented.
+"
 ```
 The external reviews arrive asynchronously — their comments will be picked up by **address-pr-comments** in Sub-step 2B1.
 
@@ -240,7 +246,7 @@ Check **all three** review sources for remaining issues:
 | Any findings | Any | Any | **Continue** → go back to Sub-step 2A1 ∥ 2A2 |
 | APPROVE | Unresolved threads | Any | **Continue** → go back to Sub-step 2A1 ∥ 2A2 (address-pr-comments will handle them) |
 | APPROVE | None unresolved | Failing | **Continue** → go back to Sub-step 2A1 ∥ 2A2 (fix-ci-tests will handle it) |
-| — | — | — | If `iteration > 10` → **STOP — iteration limit reached** |
+| — | — | — | If `iteration > 30` → **STOP — iteration limit reached** |
 
 Log the iteration result before continuing or stopping:
 - Iteration number
@@ -331,9 +337,13 @@ Run a final verification regardless of how the loop exited:
 
 Record the final state of each dimension (self-review, external reviews, CI, Codex response).
 
-**If any verification fails** (CI failing, unresolved threads remain, unpushed commits that can't be pushed, or Codex hasn't responded to the latest review request), reset Step 2 and all its sub-steps to `pending`, and go back to **Step 2: Run the review-fix loop** for another iteration. Only proceed to Step 4 when all verifications pass.
+Track how many times Step 3 has **succeeded** (all four verifications passed) across the entire run.
+
+**If any verification fails** (CI failing, unresolved threads remain, unpushed commits that can't be pushed, or Codex hasn't responded to the latest review request), reset the success counter to 0, reset Step 2 and all its sub-steps to `pending`, and go back to **Step 2: Run the review-fix loop** for another iteration.
+
+**If all verifications pass**, increment the success counter. If this is the **5th consecutive success** of Step 3 → proceed to **Step 4**. Otherwise → reset Step 2 and all its sub-steps to `pending`, and go back to **Step 2: Run the review-fix loop** for another iteration to re-confirm stability.
 
-**Completion check:** All four verifications passed. Mark Step 3 as `completed`.
+**Completion check:** Step 3 has succeeded 5 consecutive times. Mark Step 3 as `completed`.
 
 ---
 
@@ -385,5 +395,5 @@ gh pr comment <pr-number> --body "<the summary markdown above>"
 - **Run address-pr-comments before fix-ci-tests** — 2B then 2C, sequentially, so CI fixes run on code that already incorporates review feedback.
 - **Pull before fixing** — always `git pull --rebase` before launching fix agents to avoid working on stale code.
 - **Stop early on APPROVE + CI green + no unresolved threads** — don't waste iterations if the PR is already clean.
-- **Respect the iteration limit** — hard stop at 10 to prevent infinite loops. If issues persist after 10 iterations, report what's left for the user to handle.
+- **Respect the iteration limit** — hard stop at 30 to prevent infinite loops. If issues persist after 30 iterations, report what's left for the user to handle.
 - **Use gate checks** — always call TaskList and verify prerequisites before starting a step. This prevents out-of-order execution.
diff --git a/builtins/tests/cat/cat_differential_fuzz_test.go b/builtins/tests/cat/cat_differential_fuzz_test.go
index cc7cccdd..cd11611e 100644
--- a/builtins/tests/cat/cat_differential_fuzz_test.go
+++ b/builtins/tests/cat/cat_differential_fuzz_test.go
@@ -14,8 +14,11 @@ import (
 	"os/exec"
 	"path/filepath"
 	"strings"
+	"sync/atomic"
 	"testing"
 	"time"
+
+	"github.com/DataDog/rshell/builtins/testutil"
 )
 
 // runGNUInDir runs a GNU command under LC_ALL=C.UTF-8 with its working
@@ -68,12 +71,17 @@ func FuzzCatDifferential(f *testing.F) {
 	f.Add([]byte{0xff, 0xfe, 0x00, 0x01})
 	f.Add([]byte("line1\nline2\nline3\n"))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte) {
 		if len(input) > 64*1024 {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
diff --git a/builtins/tests/cat/cat_fuzz_test.go b/builtins/tests/cat/cat_fuzz_test.go
index f97e3023..64196051 100644
--- a/builtins/tests/cat/cat_fuzz_test.go
+++ b/builtins/tests/cat/cat_fuzz_test.go
@@ -10,8 +10,11 @@ import (
 	"context"
 	"os"
 	"path/filepath"
+	"sync/atomic"
 	"testing"
 	"time"
+
+	"github.com/DataDog/rshell/builtins/testutil"
 )
 
 // FuzzCat fuzzes cat with arbitrary file content and verifies output equals input.
@@ -56,12 +59,17 @@ func FuzzCat(f *testing.F) {
 	// ELF magic bytes (binary format detection)
 	f.Add([]byte{0x7f, 'E', 'L', 'F', 0x02, 0x01, 0x01, 0x00})
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte) {
 		if len(input) > 1<<20 {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644)
 		if err != nil {
 			t.Fatal(err)
@@ -102,12 +110,17 @@ func FuzzCatNumberLines(f *testing.F) {
 	// High bytes in line
 	f.Add([]byte{0x80, 0x81, '\n'})
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte) {
 		if len(input) > 1<<20 {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644)
 		if err != nil {
 			t.Fatal(err)
@@ -146,6 +159,9 @@ func FuzzCatDisplayFlags(f *testing.F) {
 	// Surrogate / bad UTF-8 with -v
 	f.Add([]byte{0xed, 0xa0, 0x80, '\n'}, true, false, false)
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, flagV, flagE, flagT bool) {
 		if len(input) > 1<<20 {
 			return
@@ -154,7 +170,9 @@ func FuzzCatDisplayFlags(f *testing.F) {
 			return // plain cat is covered by FuzzCat
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "input.bin"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
@@ -192,12 +210,17 @@ func FuzzCatStdin(f *testing.F) {
 	f.Add([]byte{0xfc, 0x80, 0x80, 0x80, 0x80, 0xaf, '\n'})
 	f.Add([]byte("line1\r\nline2\r\n"))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte) {
 		if len(input) > 1<<20 {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		err := os.WriteFile(filepath.Join(dir, "stdin.txt"), input, 0644)
 		if err != nil {
 			t.Fatal(err)
diff --git a/builtins/tests/cut/cut_fuzz_test.go b/builtins/tests/cut/cut_fuzz_test.go
index 68944bd2..cb3409bd 100644
--- a/builtins/tests/cut/cut_fuzz_test.go
+++ b/builtins/tests/cut/cut_fuzz_test.go
@@ -11,6 +11,7 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
+	"sync/atomic"
 	"testing"
 	"time"
 	"unicode/utf8"
@@ -69,6 +70,9 @@ func FuzzCutFields(f *testing.F) {
 	f.Add([]byte("a\tb\nc\td\n"), "1")
 	f.Add([]byte("a\tb\nc\td\n"), "2")
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, fieldSpec string) {
 		if len(input) > 1<<20 {
 			return
@@ -86,7 +90,9 @@ func FuzzCutFields(f *testing.F) {
 			}
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
@@ -133,6 +139,9 @@ func FuzzCutBytes(f *testing.F) {
 	// Large position well beyond line
 	f.Add([]byte("abc\n"), "1234567890")
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, byteSpec string) {
 		if len(input) > 1<<20 {
 			return
@@ -149,7 +158,9 @@ func FuzzCutBytes(f *testing.F) {
 			}
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
@@ -186,6 +197,9 @@ func FuzzCutDelimiter(f *testing.F) {
 	// Space as delimiter
 	f.Add([]byte("a b c\n"), " ", "2")
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, delim string, fieldSpec string) {
 		if len(input) > 1<<20 {
 			return
@@ -210,7 +224,9 @@ func FuzzCutDelimiter(f *testing.F) {
 			}
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
@@ -248,6 +264,9 @@ func FuzzCutComplement(f *testing.F) {
 	// Lines at 1 MiB cap
 	f.Add(append(bytes.Repeat([]byte("a"), 1<<20-1), '\n'), "1")
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, byteSpec string) {
 		if len(input) > 1<<20 {
 			return
@@ -264,7 +283,9 @@ func FuzzCutComplement(f *testing.F) {
 			}
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
@@ -295,12 +316,17 @@ func FuzzCutStdin(f *testing.F) {
 	// Lines at 1 MiB
 	f.Add(append(bytes.Repeat([]byte("x"), 1<<20-1), '\n'))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte) {
 		if len(input) > 1<<20 {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "stdin.txt"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
diff --git a/builtins/tests/echo/echo_fuzz_test.go b/builtins/tests/echo/echo_fuzz_test.go
index 8c93d83a..3c92ec84 100644
--- a/builtins/tests/echo/echo_fuzz_test.go
+++ b/builtins/tests/echo/echo_fuzz_test.go
@@ -7,17 +7,19 @@ package echo_test
 
 import (
 	"context"
+	"sync/atomic"
 	"testing"
 	"time"
 	"unicode/utf8"
 
 	"github.com/DataDog/rshell/builtins/testutil"
-	"github.com/DataDog/rshell/interp"
 )
 
-func cmdRunCtx(ctx context.Context, t *testing.T, script, dir string) (string, string, int) {
+// fuzzRunCtx delegates to testutil.FuzzRunScriptCtx which runs the script
+// with AllowedPaths set to [dir] for proper file access in fuzz iterations.
+func fuzzRunCtx(ctx context.Context, t *testing.T, script, dir string) (string, string, int) {
 	t.Helper()
-	return testutil.RunScriptCtx(ctx, t, script, dir, interp.AllowedPaths([]string{dir}))
+	return testutil.FuzzRunScriptCtx(ctx, t, script, dir)
 }
 
 // FuzzEcho fuzzes echo with arbitrary arguments (no escape processing).
@@ -35,6 +37,9 @@ func FuzzEcho(f *testing.F) {
 	// Long argument
 	f.Add("aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa")
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, arg string) {
 		if len(arg) > 1000 {
 			return
@@ -52,11 +57,13 @@ func FuzzEcho(f *testing.F) {
 			}
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 		defer cancel()
 
-		_, _, code := cmdRunCtx(ctx, t, "echo '"+arg+"'", dir)
+		_, _, code := fuzzRunCtx(ctx, t, "echo '"+arg+"'", dir)
 		if code != 0 {
 			t.Errorf("echo unexpected exit code %d", code)
 		}
@@ -102,6 +109,9 @@ func FuzzEchoEscapes(f *testing.F) {
 	// Long sequence to stress output buffering
 	f.Add("\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n")
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, arg string) {
 		if len(arg) > 1000 {
 			return
@@ -119,11 +129,13 @@ func FuzzEchoEscapes(f *testing.F) {
 			}
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 		defer cancel()
 
-		_, _, code := cmdRunCtx(ctx, t, "echo -e '"+arg+"'", dir)
+		_, _, code := fuzzRunCtx(ctx, t, "echo -e '"+arg+"'", dir)
 		if code != 0 {
 			t.Errorf("echo -e unexpected exit code %d", code)
 		}
@@ -139,6 +151,9 @@ func FuzzEchoFlagInteraction(f *testing.F) {
 	f.Add("hi\\n", false, true, true)     // -e -E: -E wins (last)
 	f.Add("hi\\n", true, true, false)     // -n -e
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, arg string, flagN, flagE, flagBigE bool) {
 		if len(arg) > 500 {
 			return
@@ -170,11 +185,13 @@ func FuzzEchoFlagInteraction(f *testing.F) {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 		defer cancel()
 
-		_, _, code := cmdRunCtx(ctx, t, "echo"+flags+" '"+arg+"'", dir)
+		_, _, code := fuzzRunCtx(ctx, t, "echo"+flags+" '"+arg+"'", dir)
 		if code != 0 {
 			t.Errorf("echo%s unexpected exit code %d", flags, code)
 		}
diff --git a/builtins/tests/grep/grep_fuzz_test.go b/builtins/tests/grep/grep_fuzz_test.go
index 7518b47b..c8dfeb94 100644
--- a/builtins/tests/grep/grep_fuzz_test.go
+++ b/builtins/tests/grep/grep_fuzz_test.go
@@ -11,6 +11,7 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
+	"sync/atomic"
 	"testing"
 	"time"
 	"unicode/utf8"
@@ -53,6 +54,9 @@ func FuzzGrepFileContent(f *testing.F) {
 	// Multibyte content
 	f.Add([]byte("héllo\nmünchen\n"), "l")
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, pattern string) {
 		if len(input) > 1<<20 {
 			return
@@ -76,7 +80,9 @@ func FuzzGrepFileContent(f *testing.F) {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644)
 		if err != nil {
 			t.Fatal(err)
@@ -122,6 +128,9 @@ func FuzzGrepPatterns(f *testing.F) {
 	// Very long pattern
 	f.Add([]byte("aaaa\n"), "a{1,4}")
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, pattern string) {
 		if len(input) > 1<<20 {
 			return
@@ -145,7 +154,9 @@ func FuzzGrepPatterns(f *testing.F) {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
@@ -172,12 +183,17 @@ func FuzzGrepStdin(f *testing.F) {
 	f.Add([]byte("line1\r\nline2\r\n"))
 	f.Add(append(bytes.Repeat([]byte("a"), 1<<20-1), '\n'))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte) {
 		if len(input) > 1<<20 {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		err := os.WriteFile(filepath.Join(dir, "stdin.txt"), input, 0644)
 		if err != nil {
 			t.Fatal(err)
@@ -224,6 +240,9 @@ func FuzzGrepFixedStrings(f *testing.F) {
 	// Near 1 MiB line cap (CVE-2012-5667 was 2^31; we test our 1 MiB boundary)
 	f.Add(append(bytes.Repeat([]byte("a"), 1<<20-1), '\n'), "a")
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, pattern string) {
 		if len(input) > 1<<20 {
 			return
@@ -244,7 +263,9 @@ func FuzzGrepFixedStrings(f *testing.F) {
 			}
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
@@ -279,6 +300,9 @@ func FuzzGrepFlags(f *testing.F) {
 	// Binary content
 	f.Add([]byte{0xff, 0xfe, '\n'}, true, false, false, false, int64(0), int64(0))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, caseInsensitive, invertMatch, countOnly, quiet bool, afterCtx, beforeCtx int64) {
 		if len(input) > 1<<20 {
 			return
@@ -290,7 +314,9 @@ func FuzzGrepFlags(f *testing.F) {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644)
 		if err != nil {
 			t.Fatal(err)
diff --git a/builtins/tests/head/head_differential_fuzz_test.go b/builtins/tests/head/head_differential_fuzz_test.go
index 724d2ec7..d6e2e67b 100644
--- a/builtins/tests/head/head_differential_fuzz_test.go
+++ b/builtins/tests/head/head_differential_fuzz_test.go
@@ -15,8 +15,11 @@ import (
 	"os/exec"
 	"path/filepath"
 	"strings"
+	"sync/atomic"
 	"testing"
 	"time"
+
+	"github.com/DataDog/rshell/builtins/testutil"
 )
 
 // runGNUInDir runs a GNU command under LC_ALL=C.UTF-8 with its working
@@ -69,6 +72,9 @@ func FuzzHeadDifferentialLines(f *testing.F) {
 	f.Add([]byte("single line\n"), int64(1))
 	f.Add([]byte("a\nb\nc\nd\ne\n"), int64(3))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, n int64) {
 		if len(input) > 64*1024 {
 			return
@@ -77,7 +83,9 @@ func FuzzHeadDifferentialLines(f *testing.F) {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
@@ -121,6 +129,9 @@ func FuzzHeadDifferentialBytes(f *testing.F) {
 	f.Add([]byte("hello world\n"), int64(5))
 	f.Add([]byte("abcdef\n"), int64(6))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, n int64) {
 		if len(input) > 64*1024 {
 			return
@@ -129,7 +140,9 @@ func FuzzHeadDifferentialBytes(f *testing.F) {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
diff --git a/builtins/tests/head/head_fuzz_test.go b/builtins/tests/head/head_fuzz_test.go
index 05cf3c1a..a2234a05 100644
--- a/builtins/tests/head/head_fuzz_test.go
+++ b/builtins/tests/head/head_fuzz_test.go
@@ -12,8 +12,11 @@ import (
 	"os"
 	"path/filepath"
 	"strings"
+	"sync/atomic"
 	"testing"
 	"time"
+
+	"github.com/DataDog/rshell/builtins/testutil"
 )
 
 // FuzzHeadLines fuzzes head -n N with arbitrary file content.
@@ -49,6 +52,9 @@ func FuzzHeadLines(f *testing.F) {
 	// No trailing newline on last output line
 	f.Add([]byte("line1\nline2"), int64(2))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, n int64) {
 		if len(input) > 1<<20 {
 			return
@@ -60,7 +66,9 @@ func FuzzHeadLines(f *testing.F) {
 			n = 10000
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644)
 		if err != nil {
 			t.Fatal(err)
@@ -108,6 +116,9 @@ func FuzzHeadBytes(f *testing.F) {
 	// CRLF
 	f.Add([]byte("a\r\nb\r\n"), int64(3))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, n int64) {
 		if len(input) > 1<<20 {
 			return
@@ -119,7 +130,9 @@ func FuzzHeadBytes(f *testing.F) {
 			n = 10000
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644)
 		if err != nil {
 			t.Fatal(err)
@@ -154,6 +167,9 @@ func FuzzHeadStdin(f *testing.F) {
 	f.Add([]byte{0xfc, 0x80, 0x80, 0x80, 0x80, 0xaf, '\n'}, int64(1))
 	f.Add([]byte("line1\r\nline2\r\n"), int64(1))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, n int64) {
 		if len(input) > 1<<20 {
 			return
@@ -165,7 +181,9 @@ func FuzzHeadStdin(f *testing.F) {
 			n = 10000
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		err := os.WriteFile(filepath.Join(dir, "stdin.txt"), input, 0644)
 		if err != nil {
 			t.Fatal(err)
diff --git a/builtins/tests/ls/ls_fuzz_test.go b/builtins/tests/ls/ls_fuzz_test.go
index a8c5a0cf..3fd51484 100644
--- a/builtins/tests/ls/ls_fuzz_test.go
+++ b/builtins/tests/ls/ls_fuzz_test.go
@@ -9,6 +9,7 @@ import (
 	"context"
 	"os"
 	"path/filepath"
+	"sync/atomic"
 	"testing"
 	"time"
 	"unicode/utf8"
@@ -48,6 +49,9 @@ func FuzzLsFlags(f *testing.F) {
 	f.Add("README.md", false, false, false, false, false)
 	f.Add("Makefile", false, false, false, false, false)
 
+	tmpRoot := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, filename string, flagL, flagA, flagR, flagS, flagF bool) {
 		if len(filename) == 0 || len(filename) > 100 {
 			return
@@ -66,7 +70,8 @@ func FuzzLsFlags(f *testing.F) {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, tmpRoot, &counter)
+		defer cleanup()
 		if err := os.WriteFile(filepath.Join(dir, filename), []byte("content"), 0644); err != nil {
 			// Some filenames may be invalid on the OS.
 			return
@@ -103,23 +108,28 @@ func FuzzLsFlags(f *testing.F) {
 // Edge cases: maxRecursionDepth=256 (depth 255 is last valid, 256 should error),
 // empty subdirectories, hidden subdirectories.
 func FuzzLsRecursive(f *testing.F) {
+	f.Add(int64(0))
 	f.Add(int64(1))
 	f.Add(int64(3))
 	f.Add(int64(5))
-	// Near recursion depth limit (maxRecursionDepth=256)
-	f.Add(int64(254))
-	f.Add(int64(255))
-	f.Add(int64(256))
-	f.Add(int64(257))
-	// Zero and negative handled by guard
-	f.Add(int64(0))
+	f.Add(int64(10))
+	// Note: maxRecursionDepth=256 boundary is tested in ls_pentest_test.go.
+	// Fuzz seeds above 10 are excluded because nested "sub" directories
+	// exceed OS max path length well before reaching depth 256.
+
+	tmpRoot := f.TempDir()
+	var counter atomic.Int64
 
 	f.Fuzz(func(t *testing.T, depth int64) {
+		// Cap at 10 to avoid hitting OS max path length (creating 256+ nested
+		// "sub" directories exceeds filesystem limits on most platforms).
+		// The maxRecursionDepth=256 limit is tested in ls_pentest_test.go instead.
 		if depth < 0 || depth > 10 {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, tmpRoot, &counter)
+		defer cleanup()
 		current := dir
 		for i := int64(0); i < depth; i++ {
 			subdir := filepath.Join(current, "sub")
@@ -166,13 +176,17 @@ func FuzzLsHumanReadable(f *testing.F) {
 	// Negative size (shouldn't happen but check robustness)
 	f.Add(int64(512))
 
+	tmpRoot := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, fileSize int64) {
 		// Clamp to 1 MiB to avoid slow file creation.
 		if fileSize < 0 || fileSize > 1<<20 {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, tmpRoot, &counter)
+		defer cleanup()
 		// Create a file with the specified size using Truncate.
 		fpath := filepath.Join(dir, "testfile.bin")
 		fh, err := os.Create(fpath)
@@ -210,8 +224,12 @@ func FuzzLsMultipleFiles(f *testing.F) {
 	f.Add(true, false, false, true) // -lS
 	f.Add(true, false, true, false) // -lt
 
+	tmpRoot := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, flagL, flagA, flagT, flagS bool) {
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, tmpRoot, &counter)
+		defer cleanup()
 
 		// Create a mix of files and a subdirectory.
 		files := []struct {
diff --git a/builtins/tests/strings_cmd/strings_fuzz_test.go b/builtins/tests/strings_cmd/strings_fuzz_test.go
index b07daa82..a2f6decc 100644
--- a/builtins/tests/strings_cmd/strings_fuzz_test.go
+++ b/builtins/tests/strings_cmd/strings_fuzz_test.go
@@ -11,6 +11,7 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
+	"sync/atomic"
 	"testing"
 	"time"
 
@@ -76,12 +77,17 @@ func FuzzStrings(f *testing.F) {
 	// PDF magic with printable sequences inside
 	f.Add([]byte("%PDF-1.4\x00\x00\x00binary\x00more text here\x00"))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte) {
 		if len(input) > 1<<20 {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "input.bin"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
@@ -120,6 +126,9 @@ func FuzzStringsMinLen(f *testing.F) {
 	// Tab as printable (contributes to sequence length)
 	f.Add([]byte("ab\tcd\x00"), int64(4))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, minLen int64) {
 		if len(input) > 1<<20 {
 			return
@@ -128,7 +137,9 @@ func FuzzStringsMinLen(f *testing.F) {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "input.bin"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
@@ -150,13 +161,13 @@ func FuzzStringsRadix(f *testing.F) {
 	f.Add([]byte("hello\x00world\x00text\n"), "o")
 	f.Add([]byte("hello\x00world\x00text\n"), "d")
 	f.Add([]byte("hello\x00world\x00text\n"), "x")
-	// Large offset: test 7-char field formatting
-	// At offset >= 8388608 (octal 40000000), octal offset exceeds 7 chars
-	f.Add(append(bytes.Repeat([]byte{0x00}, 8388608), []byte("hello")...), "o")
-	// Offset at decimal 9999999 (7 chars), 10000000 (8 chars — overflows field)
-	f.Add(append(bytes.Repeat([]byte{0x00}, 9999995), []byte("abcde")...), "d")
-	// Hex offset boundary: 0xfffffff = 268435455 (8 hex chars)
-	f.Add(append(bytes.Repeat([]byte{0x00}, 16), []byte("hello")...), "x")
+	// Large offset: test multi-char field formatting at ~1 MiB boundary.
+	// Octal 7777777 = 2097151 — seed just under 2 MiB exercises 7-digit octal.
+	f.Add(append(bytes.Repeat([]byte{0x00}, 1<<20), []byte("hello")...), "o")
+	// Decimal offset at ~1 MiB boundary (7-digit decimal: 1048576+)
+	f.Add(append(bytes.Repeat([]byte{0x00}, 1<<20), []byte("abcde")...), "d")
+	// Hex offset at ~1 MiB boundary (0x100000 = 6 hex digits)
+	f.Add(append(bytes.Repeat([]byte{0x00}, 1<<20), []byte("hello")...), "x")
 	// Empty input
 	f.Add([]byte{}, "d")
 	// All non-printable (no output)
@@ -164,15 +175,21 @@ func FuzzStringsRadix(f *testing.F) {
 	// Multiple strings with increasing offsets
 	f.Add([]byte("hello\x00world\x00foo\x00bar\x00"), "d")
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, radix string) {
-		if len(input) > 1<<20 {
+		// Allow up to 2 MiB so the large-offset corpus seeds (~1 MiB) execute.
+		if len(input) > 2<<20 {
 			return
 		}
 		if radix != "o" && radix != "d" && radix != "x" {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "input.bin"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
@@ -200,12 +217,17 @@ func FuzzStringsStdin(f *testing.F) {
 	// Chunk boundary
 	f.Add(append(bytes.Repeat([]byte("a"), 32*1024-1), 0x00))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte) {
 		if len(input) > 1<<20 {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "stdin.bin"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
diff --git a/builtins/tests/tail/helpers_test.go b/builtins/tests/tail/helpers_test.go
index b8c88401..115dbda2 100644
--- a/builtins/tests/tail/helpers_test.go
+++ b/builtins/tests/tail/helpers_test.go
@@ -14,6 +14,7 @@ import (
 
 	"mvdan.cc/sh/v3/syntax"
 
+	"github.com/DataDog/rshell/builtins/testutil"
 	"github.com/DataDog/rshell/interp"
 )
 
@@ -51,3 +52,10 @@ func cmdRunCtx(ctx context.Context, t *testing.T, script, dir string) (string, s
 	t.Helper()
 	return runScriptCtx(ctx, t, script, dir, interp.AllowedPaths([]string{dir}))
 }
+
+// fuzzRunCtx delegates to testutil.FuzzRunScriptCtx which runs the script
+// with AllowedPaths set to [dir] for proper file access in fuzz iterations.
+func fuzzRunCtx(ctx context.Context, t *testing.T, script, dir string) (string, string, int) {
+	t.Helper()
+	return testutil.FuzzRunScriptCtx(ctx, t, script, dir)
+}
diff --git a/builtins/tests/tail/tail_differential_fuzz_test.go b/builtins/tests/tail/tail_differential_fuzz_test.go
index 3d6dd2f9..0c43db99 100644
--- a/builtins/tests/tail/tail_differential_fuzz_test.go
+++ b/builtins/tests/tail/tail_differential_fuzz_test.go
@@ -15,8 +15,11 @@ import (
 	"os/exec"
 	"path/filepath"
 	"strings"
+	"sync/atomic"
 	"testing"
 	"time"
+
+	"github.com/DataDog/rshell/builtins/testutil"
 )
 
 // runGNUInDir runs a GNU command under LC_ALL=C.UTF-8 with its working
@@ -70,6 +73,9 @@ func FuzzTailDifferential(f *testing.F) {
 	f.Add([]byte("a\nb\nc\nd\ne\n"), int64(3))
 	f.Add(bytes.Repeat([]byte("line\n"), 20), int64(5))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, n int64) {
 		if len(input) > 64*1024 {
 			return
@@ -78,7 +84,9 @@ func FuzzTailDifferential(f *testing.F) {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
diff --git a/builtins/tests/tail/tail_fuzz_test.go b/builtins/tests/tail/tail_fuzz_test.go
index 579c5366..fcb89b9a 100644
--- a/builtins/tests/tail/tail_fuzz_test.go
+++ b/builtins/tests/tail/tail_fuzz_test.go
@@ -12,8 +12,11 @@ import (
 	"os"
 	"path/filepath"
 	"strings"
+	"sync/atomic"
 	"testing"
 	"time"
+
+	"github.com/DataDog/rshell/builtins/testutil"
 )
 
 // FuzzTailLines fuzzes tail -n N with arbitrary file content.
@@ -49,6 +52,9 @@ func FuzzTailLines(f *testing.F) {
 	// Many blank lines (stress ring buffer)
 	f.Add(bytes.Repeat([]byte("\n"), 1000), int64(5))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, n int64) {
 		if len(input) > 1<<20 {
 			return
@@ -60,7 +66,9 @@ func FuzzTailLines(f *testing.F) {
 			n = 10000
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644)
 		if err != nil {
 			t.Fatal(err)
@@ -69,7 +77,7 @@ func FuzzTailLines(f *testing.F) {
 		ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 		defer cancel()
 
-		stdout, _, code := cmdRunCtx(ctx, t, fmt.Sprintf("tail -n %d input.txt", n), dir)
+		stdout, _, code := fuzzRunCtx(ctx, t, fmt.Sprintf("tail -n %d input.txt", n), dir)
 		if code != 0 && code != 1 {
 			t.Errorf("tail -n %d unexpected exit code %d", n, code)
 		}
@@ -106,6 +114,9 @@ func FuzzTailBytes(f *testing.F) {
 	// Chunk boundary (32 KiB)
 	f.Add(bytes.Repeat([]byte("z"), 32*1024+1), int64(1))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, n int64) {
 		if len(input) > 1<<20 {
 			return
@@ -117,7 +128,9 @@ func FuzzTailBytes(f *testing.F) {
 			n = 10000
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644)
 		if err != nil {
 			t.Fatal(err)
@@ -126,7 +139,7 @@ func FuzzTailBytes(f *testing.F) {
 		ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 		defer cancel()
 
-		stdout, _, code := cmdRunCtx(ctx, t, fmt.Sprintf("tail -c %d input.txt", n), dir)
+		stdout, _, code := fuzzRunCtx(ctx, t, fmt.Sprintf("tail -c %d input.txt", n), dir)
 		if code != 0 && code != 1 {
 			t.Errorf("tail -c %d unexpected exit code %d", n, code)
 		}
@@ -153,6 +166,9 @@ func FuzzTailStdin(f *testing.F) {
 	f.Add([]byte{0xfc, 0x80, 0x80, 0x80, 0x80, 0xaf, '\n'}, int64(1))
 	f.Add([]byte("line1\r\nline2\r\n"), int64(1))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, n int64) {
 		if len(input) > 1<<20 {
 			return
@@ -164,7 +180,9 @@ func FuzzTailStdin(f *testing.F) {
 			n = 10000
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		err := os.WriteFile(filepath.Join(dir, "stdin.txt"), input, 0644)
 		if err != nil {
 			t.Fatal(err)
@@ -173,7 +191,7 @@ func FuzzTailStdin(f *testing.F) {
 		ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 		defer cancel()
 
-		_, _, code := cmdRunCtx(ctx, t, fmt.Sprintf("tail -n %d < stdin.txt", n), dir)
+		_, _, code := fuzzRunCtx(ctx, t, fmt.Sprintf("tail -n %d < stdin.txt", n), dir)
 		if code != 0 && code != 1 {
 			t.Errorf("tail stdin unexpected exit code %d", code)
 		}
@@ -200,6 +218,9 @@ func FuzzTailLinesOffset(f *testing.F) {
 	// CRLF
 	f.Add([]byte("a\r\nb\r\nc\r\n"), int64(2))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, n int64) {
 		if len(input) > 1<<20 {
 			return
@@ -211,7 +232,9 @@ func FuzzTailLinesOffset(f *testing.F) {
 			n = 10000
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644)
 		if err != nil {
 			t.Fatal(err)
@@ -220,7 +243,7 @@ func FuzzTailLinesOffset(f *testing.F) {
 		ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 		defer cancel()
 
-		_, _, code := cmdRunCtx(ctx, t, fmt.Sprintf("tail -n +%d input.txt", n), dir)
+		_, _, code := fuzzRunCtx(ctx, t, fmt.Sprintf("tail -n +%d input.txt", n), dir)
 		if code != 0 && code != 1 {
 			t.Errorf("tail -n +%d unexpected exit code %d", n, code)
 		}
@@ -244,6 +267,9 @@ func FuzzTailBytesOffset(f *testing.F) {
 	// Binary content
 	f.Add([]byte{0x00, 0x01, 0x02, 0xff, 0xfe}, int64(2))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, n int64) {
 		if len(input) > 1<<20 {
 			return
@@ -255,7 +281,9 @@ func FuzzTailBytesOffset(f *testing.F) {
 			n = 10000
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644)
 		if err != nil {
 			t.Fatal(err)
@@ -264,7 +292,7 @@ func FuzzTailBytesOffset(f *testing.F) {
 		ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 		defer cancel()
 
-		_, _, code := cmdRunCtx(ctx, t, fmt.Sprintf("tail -c +%d input.txt", n), dir)
+		_, _, code := fuzzRunCtx(ctx, t, fmt.Sprintf("tail -c +%d input.txt", n), dir)
 		if code != 0 && code != 1 {
 			t.Errorf("tail -c +%d unexpected exit code %d", n, code)
 		}
diff --git a/builtins/tests/testcmd/testcmd_fuzz_test.go b/builtins/tests/testcmd/testcmd_fuzz_test.go
index 9b902a62..96ef1aea 100644
--- a/builtins/tests/testcmd/testcmd_fuzz_test.go
+++ b/builtins/tests/testcmd/testcmd_fuzz_test.go
@@ -10,6 +10,7 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
+	"sync/atomic"
 	"testing"
 	"time"
 	"unicode/utf8"
@@ -152,17 +153,26 @@ func FuzzTestFileOps(f *testing.F) {
 	// Regular file test on non-existent (should be false)
 	f.Add("-f", false)
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, op string, createFile bool) {
 		switch op {
 		case "-e", "-f", "-d", "-s", "-r", "-w", "-x", "-h", "-L", "-p":
 		default:
 			return
 		}
+		// Each iteration gets its own subdirectory to avoid races between
+		// parallel fuzz workers operating on the same file. We use a manual
+		// counter instead of t.TempDir() to avoid the per-iteration cleanup
+		// overhead that causes "context deadline exceeded" on CI.
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
 
-		dir := t.TempDir()
 		target := "testfile.txt"
+		targetPath := filepath.Join(dir, target)
 		if createFile {
-			if err := os.WriteFile(filepath.Join(dir, target), []byte("content"), 0644); err != nil {
+			if err := os.WriteFile(targetPath, []byte("content"), 0644); err != nil {
 				t.Fatal(err)
 			}
 		}
@@ -261,12 +271,28 @@ func FuzzTestNesting(f *testing.F) {
 	f.Add("1 -eq 1 -o 1 -eq 2 -a 2 -eq 2")
 
 	f.Fuzz(func(t *testing.T, expr string) {
-		if len(expr) > 200 {
+		// Keep expressions short to avoid slow evaluation on CI where
+		// fuzz workers have limited CPU; long expressions with many
+		// tokens can cause the shell interpreter to exceed the
+		// per-iteration timeout, leading to "context deadline exceeded".
+		if len(expr) > 60 {
 			return
 		}
 		if !utf8.ValidString(expr) {
 			return
 		}
+		// Limit the number of space-separated tokens to cap nesting
+		// depth and keep evaluation fast on CI. The longest seed has
+		// 11 tokens so this covers all corpus entries.
+		tokens := 0
+		for i := 0; i < len(expr); i++ {
+			if expr[i] == ' ' {
+				tokens++
+				if tokens > 11 {
+					return
+				}
+			}
+		}
 		for _, c := range expr {
 			// Filter shell metacharacters that would be interpreted by the shell
 			// parser rather than passed to the test builtin.
diff --git a/builtins/tests/uniq/uniq_fuzz_test.go b/builtins/tests/uniq/uniq_fuzz_test.go
index c4f5cc43..4c875c31 100644
--- a/builtins/tests/uniq/uniq_fuzz_test.go
+++ b/builtins/tests/uniq/uniq_fuzz_test.go
@@ -11,16 +11,18 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
+	"sync/atomic"
 	"testing"
 	"time"
 
 	"github.com/DataDog/rshell/builtins/testutil"
-	"github.com/DataDog/rshell/interp"
 )
 
-func cmdRunCtx(ctx context.Context, t *testing.T, script, dir string) (string, string, int) {
+// fuzzRunCtx delegates to testutil.FuzzRunScriptCtx which runs the script
+// with AllowedPaths set to [dir] for proper file access in fuzz iterations.
+func fuzzRunCtx(ctx context.Context, t *testing.T, script, dir string) (string, string, int) {
 	t.Helper()
-	return testutil.RunScriptCtx(ctx, t, script, dir, interp.AllowedPaths([]string{dir}))
+	return testutil.FuzzRunScriptCtx(ctx, t, script, dir)
 }
 
 // FuzzUniq fuzzes uniq with arbitrary file content.
@@ -58,12 +60,17 @@ func FuzzUniq(f *testing.F) {
 	f.Add([]byte("a\r\na\r\n"))
 	f.Add([]byte("a\r\na\n")) // CRLF vs LF — how are these compared?
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte) {
 		if len(input) > 1<<20 {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
@@ -71,7 +78,7 @@ func FuzzUniq(f *testing.F) {
 		ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 		defer cancel()
 
-		_, _, code := cmdRunCtx(ctx, t, "uniq input.txt", dir)
+		_, _, code := fuzzRunCtx(ctx, t, "uniq input.txt", dir)
 		if code != 0 && code != 1 {
 			t.Errorf("uniq unexpected exit code %d", code)
 		}
@@ -92,12 +99,17 @@ func FuzzUniqCount(f *testing.F) {
 	// CRLF
 	f.Add([]byte("a\r\na\r\nb\r\n"))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte) {
 		if len(input) > 1<<20 {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
@@ -105,7 +117,7 @@ func FuzzUniqCount(f *testing.F) {
 		ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 		defer cancel()
 
-		_, _, code := cmdRunCtx(ctx, t, "uniq -c input.txt", dir)
+		_, _, code := fuzzRunCtx(ctx, t, "uniq -c input.txt", dir)
 		if code != 0 && code != 1 {
 			t.Errorf("uniq -c unexpected exit code %d", code)
 		}
@@ -135,6 +147,9 @@ func FuzzUniqFlags(f *testing.F) {
 	// -d: only print duplicate lines
 	f.Add([]byte("a\na\nb\nc\nc\n"), true, false, false, false, int64(0), int64(0), int64(0))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte, repeated, ignoreCase, unique, nulDelim bool, skipFields, skipChars, checkChars int64) {
 		if len(input) > 1<<20 {
 			return
@@ -149,7 +164,9 @@ func FuzzUniqFlags(f *testing.F) {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
@@ -180,7 +197,7 @@ func FuzzUniqFlags(f *testing.F) {
 		ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 		defer cancel()
 
-		_, _, code := cmdRunCtx(ctx, t, "uniq"+flags+" input.txt", dir)
+		_, _, code := fuzzRunCtx(ctx, t, "uniq"+flags+" input.txt", dir)
 		if code != 0 && code != 1 {
 			t.Errorf("uniq%s unexpected exit code %d", flags, code)
 		}
@@ -195,12 +212,17 @@ func FuzzUniqStdin(f *testing.F) {
 	f.Add([]byte{0xfc, 0x80, 0x80, '\n', 0xfc, 0x80, 0x80, '\n'})
 	f.Add([]byte("line1\r\nline1\r\nline2\r\n"))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte) {
 		if len(input) > 1<<20 {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "stdin.txt"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
@@ -208,7 +230,7 @@ func FuzzUniqStdin(f *testing.F) {
 		ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 		defer cancel()
 
-		_, _, code := cmdRunCtx(ctx, t, "uniq < stdin.txt", dir)
+		_, _, code := fuzzRunCtx(ctx, t, "uniq < stdin.txt", dir)
 		if code != 0 && code != 1 {
 			t.Errorf("uniq stdin unexpected exit code %d", code)
 		}
diff --git a/builtins/tests/wc/wc_differential_fuzz_test.go b/builtins/tests/wc/wc_differential_fuzz_test.go
index 47e8c603..a120932b 100644
--- a/builtins/tests/wc/wc_differential_fuzz_test.go
+++ b/builtins/tests/wc/wc_differential_fuzz_test.go
@@ -14,8 +14,11 @@ import (
 	"os/exec"
 	"path/filepath"
 	"strings"
+	"sync/atomic"
 	"testing"
 	"time"
+
+	"github.com/DataDog/rshell/builtins/testutil"
 )
 
 // runGNUInDir runs a GNU command under LC_ALL=C.UTF-8 with its working
@@ -68,12 +71,17 @@ func FuzzWcDifferentialLines(f *testing.F) {
 	f.Add([]byte("single line\n"))
 	f.Add(bytes.Repeat([]byte("x\n"), 100))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte) {
 		if len(input) > 64*1024 {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
@@ -116,12 +124,17 @@ func FuzzWcDifferentialWords(f *testing.F) {
 	f.Add([]byte("word"))
 	f.Add(bytes.Repeat([]byte("a b "), 50))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte) {
 		if len(input) > 64*1024 {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
@@ -163,12 +176,17 @@ func FuzzWcDifferentialBytes(f *testing.F) {
 	f.Add(bytes.Repeat([]byte("x"), 100))
 	f.Add([]byte("\n\n\n"))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte) {
 		if len(input) > 64*1024 {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		if err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644); err != nil {
 			t.Fatal(err)
 		}
diff --git a/builtins/tests/wc/wc_fuzz_test.go b/builtins/tests/wc/wc_fuzz_test.go
index 3d6b35dd..9c267b1e 100644
--- a/builtins/tests/wc/wc_fuzz_test.go
+++ b/builtins/tests/wc/wc_fuzz_test.go
@@ -10,8 +10,11 @@ import (
 	"context"
 	"os"
 	"path/filepath"
+	"sync/atomic"
 	"testing"
 	"time"
+
+	"github.com/DataDog/rshell/builtins/testutil"
 )
 
 // FuzzWc fuzzes wc (default mode: lines, words, bytes) with arbitrary file content.
@@ -51,12 +54,17 @@ func FuzzWc(f *testing.F) {
 	// Long line (tests -L max-line-length tracking)
 	f.Add(append(bytes.Repeat([]byte("a"), 1000), '\n'))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte) {
 		if len(input) > 1<<20 {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644)
 		if err != nil {
 			t.Fatal(err)
@@ -84,12 +92,17 @@ func FuzzWcLines(f *testing.F) {
 	f.Add([]byte{0xfc, 0x80, 0x80, 0x80, 0x80, 0xaf, '\n'})
 	f.Add(bytes.Repeat([]byte("a\n"), 10000))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte) {
 		if len(input) > 1<<20 {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644)
 		if err != nil {
 			t.Fatal(err)
@@ -115,12 +128,17 @@ func FuzzWcBytes(f *testing.F) {
 	f.Add([]byte{0x00, 0x01, 0x02, 0xff, 0xfe})
 	f.Add([]byte{0xfc, 0x80, 0x80, 0x80, 0x80, 0xaf})
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte) {
 		if len(input) > 1<<20 {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644)
 		if err != nil {
 			t.Fatal(err)
@@ -153,12 +171,17 @@ func FuzzWcChars(f *testing.F) {
 	f.Add([]byte{})
 	f.Add([]byte("no newline"))
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte) {
 		if len(input) > 1<<20 {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		err := os.WriteFile(filepath.Join(dir, "input.txt"), input, 0644)
 		if err != nil {
 			t.Fatal(err)
@@ -185,12 +208,17 @@ func FuzzWcStdin(f *testing.F) {
 	f.Add([]byte("héllo\n"))
 	f.Add([]byte{0xfc, 0x80, 0x80, 0x80, 0x80, 0xaf, '\n'})
 
+	baseDir := f.TempDir()
+	var counter atomic.Int64
+
 	f.Fuzz(func(t *testing.T, input []byte) {
 		if len(input) > 1<<20 {
 			return
 		}
 
-		dir := t.TempDir()
+		dir, cleanup := testutil.FuzzIterDir(t, baseDir, &counter)
+		defer cleanup()
+
 		err := os.WriteFile(filepath.Join(dir, "stdin.txt"), input, 0644)
 		if err != nil {
 			t.Fatal(err)
diff --git a/builtins/testutil/testutil.go b/builtins/testutil/testutil.go
index bed6dead..151206aa 100644
--- a/builtins/testutil/testutil.go
+++ b/builtins/testutil/testutil.go
@@ -10,8 +10,12 @@ import (
 	"bytes"
 	"context"
 	"errors"
+	"fmt"
 	"io"
+	"os"
+	"path/filepath"
 	"strings"
+	"sync/atomic"
 	"testing"
 
 	"github.com/stretchr/testify/require"
@@ -79,6 +83,33 @@ func RunScriptCtx(ctx context.Context, t testing.TB, script, dir string, opts ..
 	return outBuf.String(), errBuf.String(), exitCode
 }
 
+// FuzzIterDir creates an isolated per-iteration subdirectory under baseDir,
+// using counter to generate a unique name. It returns the directory path and
+// a cleanup function that removes the directory. This replaces the ~12-line
+// boilerplate pattern (atomic counter + MkdirAll + defer RemoveAll) that was
+// previously duplicated across 30+ fuzz functions.
+func FuzzIterDir(t testing.TB, baseDir string, counter *atomic.Int64) (string, func()) {
+	t.Helper()
+	n := counter.Add(1)
+	dir := filepath.Join(baseDir, fmt.Sprintf("iter%d", n))
+	if err := os.MkdirAll(dir, 0755); err != nil {
+		t.Fatal(err)
+	}
+	return dir, func() {
+		if err := os.RemoveAll(dir); err != nil && !os.IsNotExist(err) {
+			t.Errorf("cleanup %s: %v", dir, err)
+		}
+	}
+}
+
+// FuzzRunScriptCtx runs a script in fuzz mode with AllowedPaths set to [dir].
+// This ensures fuzz iterations can actually read/write files inside their
+// iteration directory rather than silently failing with permission errors.
+func FuzzRunScriptCtx(ctx context.Context, t testing.TB, script, dir string) (string, string, int) {
+	t.Helper()
+	return RunScriptCtx(ctx, t, script, dir, interp.AllowedPaths([]string{dir}))
+}
+
 // RunScript runs a shell script and returns stdout, stderr, and the exit code.
 // It accepts testing.TB so it can be used in both tests and benchmarks.
 func RunScript(t testing.TB, script, dir string, opts ...interp.RunnerOption) (string, string, int) {
diff --git a/tests/scenarios/cmd/cat/errors/is_directory.yaml b/tests/scenarios/cmd/cat/errors/is_directory.yaml
index 8b776bee..ba5725bd 100644
--- a/tests/scenarios/cmd/cat/errors/is_directory.yaml
+++ b/tests/scenarios/cmd/cat/errors/is_directory.yaml
@@ -1,3 +1,4 @@
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 description: Cat on a directory fails with "is a directory" error and exits with code 1.
 setup:
diff --git a/tests/scenarios/cmd/cat/errors/multiple_all_fail.yaml b/tests/scenarios/cmd/cat/errors/multiple_all_fail.yaml
index 3c3218f2..a8de80f3 100644
--- a/tests/scenarios/cmd/cat/errors/multiple_all_fail.yaml
+++ b/tests/scenarios/cmd/cat/errors/multiple_all_fail.yaml
@@ -1,3 +1,4 @@
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 description: When all files are missing, cat reports errors for each file.
 input:
diff --git a/tests/scenarios/cmd/cat/errors/multiple_first_fails.yaml b/tests/scenarios/cmd/cat/errors/multiple_first_fails.yaml
index f401efd9..9cdb1b81 100644
--- a/tests/scenarios/cmd/cat/errors/multiple_first_fails.yaml
+++ b/tests/scenarios/cmd/cat/errors/multiple_first_fails.yaml
@@ -1,3 +1,4 @@
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 description: When the first of multiple files fails, cat continues and processes remaining files.
 setup:
diff --git a/tests/scenarios/cmd/cat/errors/multiple_second_fails.yaml b/tests/scenarios/cmd/cat/errors/multiple_second_fails.yaml
index 9a1e31b8..9537fbdd 100644
--- a/tests/scenarios/cmd/cat/errors/multiple_second_fails.yaml
+++ b/tests/scenarios/cmd/cat/errors/multiple_second_fails.yaml
@@ -1,3 +1,4 @@
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 description: When the second of multiple files fails, the first file is output before the error.
 setup:
diff --git a/tests/scenarios/cmd/cat/errors/nonexistent_continues.yaml b/tests/scenarios/cmd/cat/errors/nonexistent_continues.yaml
index 7733384e..8d49ce3d 100644
--- a/tests/scenarios/cmd/cat/errors/nonexistent_continues.yaml
+++ b/tests/scenarios/cmd/cat/errors/nonexistent_continues.yaml
@@ -1,3 +1,4 @@
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 description: Cat failure does not stop script execution; the exit code is from the last command.
 input:
diff --git a/tests/scenarios/cmd/cat/errors/nonexistent_file.yaml b/tests/scenarios/cmd/cat/errors/nonexistent_file.yaml
index 97d696d8..b067e57b 100644
--- a/tests/scenarios/cmd/cat/errors/nonexistent_file.yaml
+++ b/tests/scenarios/cmd/cat/errors/nonexistent_file.yaml
@@ -1,3 +1,4 @@
+# skip: rshell builtin error format differs from GNU coreutils (includes Go "openat" prefix, lowercase)
 skip_assert_against_bash: true
 description: Cat on a nonexistent file prints an error to stderr and exits with code 1.
 input:
diff --git a/tests/scenarios/cmd/cat/errors/redirect_nonexistent.yaml b/tests/scenarios/cmd/cat/errors/redirect_nonexistent.yaml
index b635538f..726c7c9a 100644
--- a/tests/scenarios/cmd/cat/errors/redirect_nonexistent.yaml
+++ b/tests/scenarios/cmd/cat/errors/redirect_nonexistent.yaml
@@ -1,3 +1,4 @@
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 description: Input redirect from a nonexistent file fails at the redirect level, not cat itself, so the error has no "cat:" prefix.
 input:
diff --git a/tests/scenarios/cmd/cat/hardening/unknown_flag.yaml b/tests/scenarios/cmd/cat/hardening/unknown_flag.yaml
index ea7ec61b..ff152d19 100644
--- a/tests/scenarios/cmd/cat/hardening/unknown_flag.yaml
+++ b/tests/scenarios/cmd/cat/hardening/unknown_flag.yaml
@@ -1,5 +1,4 @@
 description: cat rejects unknown flags with exit code 1.
-skip_assert_against_bash: true
 input:
   script: |+
     cat --follow file.txt
diff --git a/tests/scenarios/cmd/cat/help/help_flag.yaml b/tests/scenarios/cmd/cat/help/help_flag.yaml
index 2f19a2d5..943cb8c0 100644
--- a/tests/scenarios/cmd/cat/help/help_flag.yaml
+++ b/tests/scenarios/cmd/cat/help/help_flag.yaml
@@ -1,4 +1,5 @@
 description: cat --help prints usage to stdout and exits 0.
+# skip: rshell builtin help output differs from GNU coreutils
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/echo/escapes/unicode_locale_independent.yaml b/tests/scenarios/cmd/echo/escapes/unicode_locale_independent.yaml
index 49cb0602..6f7c792e 100644
--- a/tests/scenarios/cmd/echo/escapes/unicode_locale_independent.yaml
+++ b/tests/scenarios/cmd/echo/escapes/unicode_locale_independent.yaml
@@ -1,4 +1,5 @@
 description: Echo -e with non-ASCII unicode is locale-independent (intentional divergence from bash which is locale-dependent).
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/echo/escapes/unicode_out_of_range.yaml b/tests/scenarios/cmd/echo/escapes/unicode_out_of_range.yaml
index 5fd98410..527a957d 100644
--- a/tests/scenarios/cmd/echo/escapes/unicode_out_of_range.yaml
+++ b/tests/scenarios/cmd/echo/escapes/unicode_out_of_range.yaml
@@ -1,4 +1,5 @@
 description: Echo -e with out-of-range Unicode \UFFFFFFFF silently drops the codepoint (intentional divergence from bash).
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/echo/escapes/unicode_surrogate.yaml b/tests/scenarios/cmd/echo/escapes/unicode_surrogate.yaml
index 6f8cec9f..347561b8 100644
--- a/tests/scenarios/cmd/echo/escapes/unicode_surrogate.yaml
+++ b/tests/scenarios/cmd/echo/escapes/unicode_surrogate.yaml
@@ -1,4 +1,5 @@
 description: Echo -e with surrogate codepoint replaces with U+FFFD (intentional security divergence from bash).
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/grep/errors/conflicting_matchers.yaml b/tests/scenarios/cmd/grep/errors/conflicting_matchers.yaml
index 89448913..fc79fb26 100644
--- a/tests/scenarios/cmd/grep/errors/conflicting_matchers.yaml
+++ b/tests/scenarios/cmd/grep/errors/conflicting_matchers.yaml
@@ -1,3 +1,4 @@
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 description: grep rejects conflicting matcher flags with exit code 2.
 input:
diff --git a/tests/scenarios/cmd/grep/errors/invalid_regex.yaml b/tests/scenarios/cmd/grep/errors/invalid_regex.yaml
index 560cdfd2..5eb8e438 100644
--- a/tests/scenarios/cmd/grep/errors/invalid_regex.yaml
+++ b/tests/scenarios/cmd/grep/errors/invalid_regex.yaml
@@ -1,3 +1,4 @@
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 description: grep with invalid regex exits 2.
 setup:
diff --git a/tests/scenarios/cmd/grep/errors/no_pattern.yaml b/tests/scenarios/cmd/grep/errors/no_pattern.yaml
index 819a2426..36943994 100644
--- a/tests/scenarios/cmd/grep/errors/no_pattern.yaml
+++ b/tests/scenarios/cmd/grep/errors/no_pattern.yaml
@@ -1,3 +1,4 @@
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 description: grep with no pattern exits 2.
 input:
diff --git a/tests/scenarios/cmd/grep/errors/nonexistent_file.yaml b/tests/scenarios/cmd/grep/errors/nonexistent_file.yaml
index 3e2de845..e28d1fb6 100644
--- a/tests/scenarios/cmd/grep/errors/nonexistent_file.yaml
+++ b/tests/scenarios/cmd/grep/errors/nonexistent_file.yaml
@@ -1,3 +1,4 @@
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 description: grep on a nonexistent file prints an error to stderr and exits with code 2.
 input:
diff --git a/tests/scenarios/cmd/grep/flags/help.yaml b/tests/scenarios/cmd/grep/flags/help.yaml
index 36457bcc..6e8e88d0 100644
--- a/tests/scenarios/cmd/grep/flags/help.yaml
+++ b/tests/scenarios/cmd/grep/flags/help.yaml
@@ -1,4 +1,5 @@
 description: grep --help prints usage information and exits with code 0.
+# skip: rshell builtin help output differs from GNU coreutils
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/grep/hardening/unknown_flag.yaml b/tests/scenarios/cmd/grep/hardening/unknown_flag.yaml
index 5f48c478..300b5cbc 100644
--- a/tests/scenarios/cmd/grep/hardening/unknown_flag.yaml
+++ b/tests/scenarios/cmd/grep/hardening/unknown_flag.yaml
@@ -1,4 +1,5 @@
 description: grep rejects unknown flags with exit code 1.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/head/errors/boolean_flag_with_argument.yaml b/tests/scenarios/cmd/head/errors/boolean_flag_with_argument.yaml
index 5c090f48..2ddc2e52 100644
--- a/tests/scenarios/cmd/head/errors/boolean_flag_with_argument.yaml
+++ b/tests/scenarios/cmd/head/errors/boolean_flag_with_argument.yaml
@@ -1,6 +1,7 @@
 # GNU head rejects --quiet=<value> and --verbose=<value> (options don't allow arguments).
 # Our error message differs from GNU head so bash comparison is skipped.
 description: head exits 1 when --quiet or --verbose are given an explicit argument.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/head/errors/negative_bytes_count.yaml b/tests/scenarios/cmd/head/errors/negative_bytes_count.yaml
index fa67bf6d..c9421d39 100644
--- a/tests/scenarios/cmd/head/errors/negative_bytes_count.yaml
+++ b/tests/scenarios/cmd/head/errors/negative_bytes_count.yaml
@@ -1,5 +1,6 @@
 # We intentionally reject negative byte counts (we do not implement -c -N elide-tail mode)
 description: head exits 1 when -c is given a negative count.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/head/errors/negative_count.yaml b/tests/scenarios/cmd/head/errors/negative_count.yaml
index 3ad71682..34241732 100644
--- a/tests/scenarios/cmd/head/errors/negative_count.yaml
+++ b/tests/scenarios/cmd/head/errors/negative_count.yaml
@@ -1,5 +1,6 @@
 # We intentionally reject negative counts (we do not implement -n -N elide-tail mode)
 description: head exits 1 when -n is given a negative count.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/head/hardening/outside_allowed_paths.yaml b/tests/scenarios/cmd/head/hardening/outside_allowed_paths.yaml
index bd0df4c5..2e011f3d 100644
--- a/tests/scenarios/cmd/head/hardening/outside_allowed_paths.yaml
+++ b/tests/scenarios/cmd/head/hardening/outside_allowed_paths.yaml
@@ -1,5 +1,6 @@
 # Per RULES.md: file access must be sandboxed via AllowedPaths
 description: head is blocked from reading files outside the allowed paths sandbox.
+# skip: sandbox path restrictions are an rshell-specific feature
 skip_assert_against_bash: true  # intentional sandbox restriction; bash/head can read /etc/passwd freely
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/basic/empty_string.yaml b/tests/scenarios/cmd/ls/basic/empty_string.yaml
index 8d9647d4..96a61eb9 100644
--- a/tests/scenarios/cmd/ls/basic/empty_string.yaml
+++ b/tests/scenarios/cmd/ls/basic/empty_string.yaml
@@ -1,4 +1,5 @@
 description: ls with an empty string operand prints an error and exits 1.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 input:
   allowed_paths: ["$DIR"]
diff --git a/tests/scenarios/cmd/ls/basic/mixed_valid_invalid.yaml b/tests/scenarios/cmd/ls/basic/mixed_valid_invalid.yaml
index 59e9a7c7..e0fdba4b 100644
--- a/tests/scenarios/cmd/ls/basic/mixed_valid_invalid.yaml
+++ b/tests/scenarios/cmd/ls/basic/mixed_valid_invalid.yaml
@@ -1,4 +1,5 @@
 description: ls with a mix of valid and invalid args lists valid ones and reports errors.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/basic/nonexistent.yaml b/tests/scenarios/cmd/ls/basic/nonexistent.yaml
index 5876aabd..0056a269 100644
--- a/tests/scenarios/cmd/ls/basic/nonexistent.yaml
+++ b/tests/scenarios/cmd/ls/basic/nonexistent.yaml
@@ -1,4 +1,5 @@
 description: ls on a nonexistent path prints an error and exits 1.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 input:
   allowed_paths: ["$DIR"]
diff --git a/tests/scenarios/cmd/ls/flags/all.yaml b/tests/scenarios/cmd/ls/flags/all.yaml
index 33827fe1..f5ff227b 100644
--- a/tests/scenarios/cmd/ls/flags/all.yaml
+++ b/tests/scenarios/cmd/ls/flags/all.yaml
@@ -1,4 +1,5 @@
 description: ls -a shows hidden files including . and ..
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/flags/all_override.yaml b/tests/scenarios/cmd/ls/flags/all_override.yaml
index 9dd6ebc5..6199a6a7 100644
--- a/tests/scenarios/cmd/ls/flags/all_override.yaml
+++ b/tests/scenarios/cmd/ls/flags/all_override.yaml
@@ -1,4 +1,5 @@
 description: ls -a -A — last flag wins, -A suppresses . and ..
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/flags/help.yaml b/tests/scenarios/cmd/ls/flags/help.yaml
index d1b9b33d..082307f8 100644
--- a/tests/scenarios/cmd/ls/flags/help.yaml
+++ b/tests/scenarios/cmd/ls/flags/help.yaml
@@ -1,4 +1,5 @@
 description: ls --help prints usage information and exits with code 0.
+# skip: rshell builtin help output differs from GNU coreutils
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/ls/flags/sort_by_size.yaml b/tests/scenarios/cmd/ls/flags/sort_by_size.yaml
index 08ee4a76..8b06607e 100644
--- a/tests/scenarios/cmd/ls/flags/sort_by_size.yaml
+++ b/tests/scenarios/cmd/ls/flags/sort_by_size.yaml
@@ -1,4 +1,5 @@
 description: ls -S sorts by file size, largest first.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/flags/sort_by_time.yaml b/tests/scenarios/cmd/ls/flags/sort_by_time.yaml
index e5758821..845ced32 100644
--- a/tests/scenarios/cmd/ls/flags/sort_by_time.yaml
+++ b/tests/scenarios/cmd/ls/flags/sort_by_time.yaml
@@ -1,4 +1,5 @@
 description: ls -t sorts by modification time, newest first.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/long_format/basic_long.yaml b/tests/scenarios/cmd/ls/long_format/basic_long.yaml
index 0c34cf9d..58f6045c 100644
--- a/tests/scenarios/cmd/ls/long_format/basic_long.yaml
+++ b/tests/scenarios/cmd/ls/long_format/basic_long.yaml
@@ -1,4 +1,5 @@
 description: ls -l shows long format with mode, size, date, and name.
+# skip: rshell ls output format differs from GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/long_format/human_readable.yaml b/tests/scenarios/cmd/ls/long_format/human_readable.yaml
index 69a5c354..264b1887 100644
--- a/tests/scenarios/cmd/ls/long_format/human_readable.yaml
+++ b/tests/scenarios/cmd/ls/long_format/human_readable.yaml
@@ -1,4 +1,5 @@
 description: ls -lh shows human-readable sizes.
+# skip: rshell ls output format differs from GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/pagination/directory_flag_with_limit.yaml b/tests/scenarios/cmd/ls/pagination/directory_flag_with_limit.yaml
index e866eb4e..651644b6 100644
--- a/tests/scenarios/cmd/ls/pagination/directory_flag_with_limit.yaml
+++ b/tests/scenarios/cmd/ls/pagination/directory_flag_with_limit.yaml
@@ -1,4 +1,5 @@
 description: ls -d --limit lists directories themselves without entering them.
+# skip: pagination flags are an rshell-specific extension not present in GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/pagination/dot_dotdot_not_consumed_by_limit.yaml b/tests/scenarios/cmd/ls/pagination/dot_dotdot_not_consumed_by_limit.yaml
index 1515490e..bb968079 100644
--- a/tests/scenarios/cmd/ls/pagination/dot_dotdot_not_consumed_by_limit.yaml
+++ b/tests/scenarios/cmd/ls/pagination/dot_dotdot_not_consumed_by_limit.yaml
@@ -1,4 +1,5 @@
 description: ls -a --limit does not count . and .. against the limit.
+# skip: pagination flags are an rshell-specific extension not present in GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/pagination/dot_dotdot_not_consumed_by_offset.yaml b/tests/scenarios/cmd/ls/pagination/dot_dotdot_not_consumed_by_offset.yaml
index d71a84e6..fe88f1f8 100644
--- a/tests/scenarios/cmd/ls/pagination/dot_dotdot_not_consumed_by_offset.yaml
+++ b/tests/scenarios/cmd/ls/pagination/dot_dotdot_not_consumed_by_offset.yaml
@@ -1,4 +1,5 @@
 description: ls -a --offset does not skip . and .. (they are synthesized after pagination).
+# skip: pagination flags are an rshell-specific extension not present in GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/pagination/limit_only.yaml b/tests/scenarios/cmd/ls/pagination/limit_only.yaml
index d581db04..3b6b8b85 100644
--- a/tests/scenarios/cmd/ls/pagination/limit_only.yaml
+++ b/tests/scenarios/cmd/ls/pagination/limit_only.yaml
@@ -1,4 +1,5 @@
 description: ls --limit caps the number of entries returned.
+# skip: pagination flags are an rshell-specific extension not present in GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/pagination/limit_with_reverse.yaml b/tests/scenarios/cmd/ls/pagination/limit_with_reverse.yaml
index e7382c63..4ded82d1 100644
--- a/tests/scenarios/cmd/ls/pagination/limit_with_reverse.yaml
+++ b/tests/scenarios/cmd/ls/pagination/limit_with_reverse.yaml
@@ -1,4 +1,5 @@
 description: ls -r --limit returns the correct number of entries in reverse sort order.
+# skip: pagination flags are an rshell-specific extension not present in GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/pagination/multi_dir_ignores_limit.yaml b/tests/scenarios/cmd/ls/pagination/multi_dir_ignores_limit.yaml
index 0315339f..f01a0adc 100644
--- a/tests/scenarios/cmd/ls/pagination/multi_dir_ignores_limit.yaml
+++ b/tests/scenarios/cmd/ls/pagination/multi_dir_ignores_limit.yaml
@@ -1,4 +1,5 @@
 description: ls with multiple directory args silently ignores --limit.
+# skip: pagination flags are an rshell-specific extension not present in GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/pagination/multi_dir_ignores_offset.yaml b/tests/scenarios/cmd/ls/pagination/multi_dir_ignores_offset.yaml
index f46d3ae4..b5538569 100644
--- a/tests/scenarios/cmd/ls/pagination/multi_dir_ignores_offset.yaml
+++ b/tests/scenarios/cmd/ls/pagination/multi_dir_ignores_offset.yaml
@@ -1,4 +1,5 @@
 description: ls with multiple directory args silently ignores --offset.
+# skip: pagination flags are an rshell-specific extension not present in GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/pagination/negative_limit_clamped.yaml b/tests/scenarios/cmd/ls/pagination/negative_limit_clamped.yaml
index 5ee4023b..85057d05 100644
--- a/tests/scenarios/cmd/ls/pagination/negative_limit_clamped.yaml
+++ b/tests/scenarios/cmd/ls/pagination/negative_limit_clamped.yaml
@@ -1,4 +1,5 @@
 description: Negative --limit is clamped to zero (shows all entries).
+# skip: pagination flags are an rshell-specific extension not present in GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/pagination/negative_offset_clamped.yaml b/tests/scenarios/cmd/ls/pagination/negative_offset_clamped.yaml
index 267ad534..994cca14 100644
--- a/tests/scenarios/cmd/ls/pagination/negative_offset_clamped.yaml
+++ b/tests/scenarios/cmd/ls/pagination/negative_offset_clamped.yaml
@@ -1,4 +1,5 @@
 description: Negative --offset is clamped to zero.
+# skip: pagination flags are an rshell-specific extension not present in GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/pagination/offset_and_limit.yaml b/tests/scenarios/cmd/ls/pagination/offset_and_limit.yaml
index 141641de..31308e7c 100644
--- a/tests/scenarios/cmd/ls/pagination/offset_and_limit.yaml
+++ b/tests/scenarios/cmd/ls/pagination/offset_and_limit.yaml
@@ -1,4 +1,5 @@
 description: ls --offset and --limit together paginate results.
+# skip: pagination flags are an rshell-specific extension not present in GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/pagination/offset_beyond_count.yaml b/tests/scenarios/cmd/ls/pagination/offset_beyond_count.yaml
index d91c4a3e..94e2f0cc 100644
--- a/tests/scenarios/cmd/ls/pagination/offset_beyond_count.yaml
+++ b/tests/scenarios/cmd/ls/pagination/offset_beyond_count.yaml
@@ -1,4 +1,5 @@
 description: ls --offset beyond entry count returns empty output.
+# skip: pagination flags are an rshell-specific extension not present in GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/pagination/offset_only.yaml b/tests/scenarios/cmd/ls/pagination/offset_only.yaml
index ffbeb394..f59ccf3c 100644
--- a/tests/scenarios/cmd/ls/pagination/offset_only.yaml
+++ b/tests/scenarios/cmd/ls/pagination/offset_only.yaml
@@ -1,4 +1,5 @@
 description: ls --offset skips entries and returns the rest.
+# skip: pagination flags are an rshell-specific extension not present in GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/pagination/recursive_ignores_limit.yaml b/tests/scenarios/cmd/ls/pagination/recursive_ignores_limit.yaml
index f8b4bfa6..b0a6746d 100644
--- a/tests/scenarios/cmd/ls/pagination/recursive_ignores_limit.yaml
+++ b/tests/scenarios/cmd/ls/pagination/recursive_ignores_limit.yaml
@@ -1,4 +1,5 @@
 description: ls -R silently ignores --limit.
+# skip: pagination flags are an rshell-specific extension not present in GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/pagination/recursive_ignores_offset.yaml b/tests/scenarios/cmd/ls/pagination/recursive_ignores_offset.yaml
index 171a8e06..3edcaf47 100644
--- a/tests/scenarios/cmd/ls/pagination/recursive_ignores_offset.yaml
+++ b/tests/scenarios/cmd/ls/pagination/recursive_ignores_offset.yaml
@@ -1,4 +1,5 @@
 description: ls -R silently ignores --offset.
+# skip: pagination flags are an rshell-specific extension not present in GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/pagination/zero_offset_zero_limit.yaml b/tests/scenarios/cmd/ls/pagination/zero_offset_zero_limit.yaml
index 23f1e4d2..c6a6f8a1 100644
--- a/tests/scenarios/cmd/ls/pagination/zero_offset_zero_limit.yaml
+++ b/tests/scenarios/cmd/ls/pagination/zero_offset_zero_limit.yaml
@@ -1,4 +1,5 @@
 description: --offset 0 --limit 0 behaves like no pagination.
+# skip: pagination flags are an rshell-specific extension not present in GNU coreutils
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/sandbox/outside_allowed_paths.yaml b/tests/scenarios/cmd/ls/sandbox/outside_allowed_paths.yaml
index bc70f890..0f5af22c 100644
--- a/tests/scenarios/cmd/ls/sandbox/outside_allowed_paths.yaml
+++ b/tests/scenarios/cmd/ls/sandbox/outside_allowed_paths.yaml
@@ -1,4 +1,5 @@
 description: ls outside allowed_paths returns permission denied.
+# skip: sandbox path restrictions are an rshell-specific feature
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/ls/sandbox/parent_traversal.yaml b/tests/scenarios/cmd/ls/sandbox/parent_traversal.yaml
index 3b07e6f9..3249b937 100644
--- a/tests/scenarios/cmd/ls/sandbox/parent_traversal.yaml
+++ b/tests/scenarios/cmd/ls/sandbox/parent_traversal.yaml
@@ -1,4 +1,5 @@
 description: ls with parent traversal outside sandbox returns permission denied.
+# skip: sandbox path restrictions are an rshell-specific feature
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/cmd/printf/errors/rejected_n_specifier.yaml b/tests/scenarios/cmd/printf/errors/rejected_n_specifier.yaml
index f9f1b601..155b93c1 100644
--- a/tests/scenarios/cmd/printf/errors/rejected_n_specifier.yaml
+++ b/tests/scenarios/cmd/printf/errors/rejected_n_specifier.yaml
@@ -6,4 +6,5 @@ expect:
   stdout: ""
   stderr: "printf: %n: not supported (security risk)\n"
   exit_code: 1
+# skip: feature is intentionally blocked in rshell for security
 skip_assert_against_bash: true
diff --git a/tests/scenarios/cmd/printf/errors/rejected_q_specifier.yaml b/tests/scenarios/cmd/printf/errors/rejected_q_specifier.yaml
index 4e0033bf..08e4763f 100644
--- a/tests/scenarios/cmd/printf/errors/rejected_q_specifier.yaml
+++ b/tests/scenarios/cmd/printf/errors/rejected_q_specifier.yaml
@@ -1,4 +1,5 @@
 description: printf rejects the %q format specifier with a clear error message.
+# skip: feature is intentionally blocked in rshell for security
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/printf/errors/rejected_v_flag.yaml b/tests/scenarios/cmd/printf/errors/rejected_v_flag.yaml
index 3ee2cadf..a284e6e7 100644
--- a/tests/scenarios/cmd/printf/errors/rejected_v_flag.yaml
+++ b/tests/scenarios/cmd/printf/errors/rejected_v_flag.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in rshell for security
 skip_assert_against_bash: true
 description: Printf rejects the -v flag which assigns to a variable in bash.
 input:
diff --git a/tests/scenarios/cmd/printf/escapes/unicode_emoji.yaml b/tests/scenarios/cmd/printf/escapes/unicode_emoji.yaml
index c91c2447..fce7600f 100644
--- a/tests/scenarios/cmd/printf/escapes/unicode_emoji.yaml
+++ b/tests/scenarios/cmd/printf/escapes/unicode_emoji.yaml
@@ -1,4 +1,5 @@
 description: Printf interprets \U for emoji Unicode code points.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # bash outputs literal \U in POSIX locale; rshell always emits UTF-8
 input:
   script: |+
diff --git a/tests/scenarios/cmd/printf/escapes/unicode_multibyte.yaml b/tests/scenarios/cmd/printf/escapes/unicode_multibyte.yaml
index 1a01a1fe..ad9e6804 100644
--- a/tests/scenarios/cmd/printf/escapes/unicode_multibyte.yaml
+++ b/tests/scenarios/cmd/printf/escapes/unicode_multibyte.yaml
@@ -1,4 +1,5 @@
 description: Printf interprets \u for multi-byte Unicode characters.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # bash outputs literal \u in POSIX locale; rshell always emits UTF-8
 input:
   script: |+
diff --git a/tests/scenarios/cmd/printf/numeric/float_overflow.yaml b/tests/scenarios/cmd/printf/numeric/float_overflow.yaml
index 139a16cd..e5b93cbe 100644
--- a/tests/scenarios/cmd/printf/numeric/float_overflow.yaml
+++ b/tests/scenarios/cmd/printf/numeric/float_overflow.yaml
@@ -1,4 +1,5 @@
 description: Float range overflow outputs inf with warning and exit 0 (Go uses +Inf for overflow).
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # Go's strconv.ParseFloat returns +Inf; C's strtod + printf outputs the actual number
 input:
   script: |+
diff --git a/tests/scenarios/cmd/printf/numeric/signed_nan.yaml b/tests/scenarios/cmd/printf/numeric/signed_nan.yaml
index db8bb3de..8946e072 100644
--- a/tests/scenarios/cmd/printf/numeric/signed_nan.yaml
+++ b/tests/scenarios/cmd/printf/numeric/signed_nan.yaml
@@ -1,4 +1,5 @@
 description: Signed NaN forms +nan and -nan are accepted (bash compat).
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # Go's math.NaN() doesn't preserve sign; bash outputs -nan for -nan
 input:
   script: |+
diff --git a/tests/scenarios/cmd/printf/shell_features/command_substitution.yaml b/tests/scenarios/cmd/printf/shell_features/command_substitution.yaml
index 3ddade23..430e7258 100644
--- a/tests/scenarios/cmd/printf/shell_features/command_substitution.yaml
+++ b/tests/scenarios/cmd/printf/shell_features/command_substitution.yaml
@@ -1,4 +1,5 @@
 description: Printf output can be captured via command substitution.
+# skip: command substitution is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/printf/specifiers/b_hex_one_digit.yaml b/tests/scenarios/cmd/printf/specifiers/b_hex_one_digit.yaml
index 58ea791b..2bc117cc 100644
--- a/tests/scenarios/cmd/printf/specifiers/b_hex_one_digit.yaml
+++ b/tests/scenarios/cmd/printf/specifiers/b_hex_one_digit.yaml
@@ -5,4 +5,5 @@ input:
 expect:
   stdout: "\x0f"
   exit_code: 0
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
diff --git a/tests/scenarios/cmd/printf/specifiers/b_octal_no_digits.yaml b/tests/scenarios/cmd/printf/specifiers/b_octal_no_digits.yaml
index 5ac00f8a..b17c3b46 100644
--- a/tests/scenarios/cmd/printf/specifiers/b_octal_no_digits.yaml
+++ b/tests/scenarios/cmd/printf/specifiers/b_octal_no_digits.yaml
@@ -5,4 +5,5 @@ input:
 expect:
   stdout: "\0x"
   exit_code: 0
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
diff --git a/tests/scenarios/cmd/printf/specifiers/float_hex_large_unsigned.yaml b/tests/scenarios/cmd/printf/specifiers/float_hex_large_unsigned.yaml
index e47c8425..ce337c09 100644
--- a/tests/scenarios/cmd/printf/specifiers/float_hex_large_unsigned.yaml
+++ b/tests/scenarios/cmd/printf/specifiers/float_hex_large_unsigned.yaml
@@ -1,4 +1,5 @@
 description: "%f handles large unsigned hex values like 0xffffffffffffffff."
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # Go float64 rounds 2^64-1 to 2^64; C printf preserves exact value
 input:
   script: |+
diff --git a/tests/scenarios/cmd/sed/edge/no_trailing_newline.yaml b/tests/scenarios/cmd/sed/edge/no_trailing_newline.yaml
index 9d192bd7..a8580fc1 100644
--- a/tests/scenarios/cmd/sed/edge/no_trailing_newline.yaml
+++ b/tests/scenarios/cmd/sed/edge/no_trailing_newline.yaml
@@ -7,6 +7,7 @@ setup:
   files:
     - path: input.txt
       content: "no newline at end"
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 input:
   allowed_paths: ["$DIR"]
diff --git a/tests/scenarios/cmd/sed/errors/address_zero.yaml b/tests/scenarios/cmd/sed/errors/address_zero.yaml
index b049f7aa..068a8276 100644
--- a/tests/scenarios/cmd/sed/errors/address_zero.yaml
+++ b/tests/scenarios/cmd/sed/errors/address_zero.yaml
@@ -1,4 +1,5 @@
 description: Line address 0 as standalone address is rejected.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/sed/errors/blocked_R_cmd.yaml b/tests/scenarios/cmd/sed/errors/blocked_R_cmd.yaml
index f5064b10..f3765be3 100644
--- a/tests/scenarios/cmd/sed/errors/blocked_R_cmd.yaml
+++ b/tests/scenarios/cmd/sed/errors/blocked_R_cmd.yaml
@@ -1,4 +1,5 @@
 description: The R command is blocked for safety.
+# skip: feature is intentionally blocked in rshell for security
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/sed/errors/blocked_W_cmd.yaml b/tests/scenarios/cmd/sed/errors/blocked_W_cmd.yaml
index 831847db..fbf95c2d 100644
--- a/tests/scenarios/cmd/sed/errors/blocked_W_cmd.yaml
+++ b/tests/scenarios/cmd/sed/errors/blocked_W_cmd.yaml
@@ -1,4 +1,5 @@
 description: The W command is blocked for safety.
+# skip: feature is intentionally blocked in rshell for security
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/sed/errors/blocked_e_flag.yaml b/tests/scenarios/cmd/sed/errors/blocked_e_flag.yaml
index 665114f5..e2366638 100644
--- a/tests/scenarios/cmd/sed/errors/blocked_e_flag.yaml
+++ b/tests/scenarios/cmd/sed/errors/blocked_e_flag.yaml
@@ -1,4 +1,5 @@
 description: The e flag in s command is blocked for safety.
+# skip: feature is intentionally blocked in rshell for security
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/sed/errors/blocked_execute.yaml b/tests/scenarios/cmd/sed/errors/blocked_execute.yaml
index 89678dcb..69d16fbe 100644
--- a/tests/scenarios/cmd/sed/errors/blocked_execute.yaml
+++ b/tests/scenarios/cmd/sed/errors/blocked_execute.yaml
@@ -1,4 +1,5 @@
 description: The e command is blocked for safety.
+# skip: feature is intentionally blocked in rshell for security
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/sed/errors/blocked_inplace.yaml b/tests/scenarios/cmd/sed/errors/blocked_inplace.yaml
index 8db911f8..e2380aca 100644
--- a/tests/scenarios/cmd/sed/errors/blocked_inplace.yaml
+++ b/tests/scenarios/cmd/sed/errors/blocked_inplace.yaml
@@ -1,4 +1,5 @@
 description: The -i flag is blocked for safety.
+# skip: feature is intentionally blocked in rshell for security
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/sed/errors/blocked_read.yaml b/tests/scenarios/cmd/sed/errors/blocked_read.yaml
index 5b069997..1e0d7db2 100644
--- a/tests/scenarios/cmd/sed/errors/blocked_read.yaml
+++ b/tests/scenarios/cmd/sed/errors/blocked_read.yaml
@@ -1,4 +1,5 @@
 description: The r command is blocked for safety.
+# skip: feature is intentionally blocked in rshell for security
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/sed/errors/blocked_write.yaml b/tests/scenarios/cmd/sed/errors/blocked_write.yaml
index 7336792c..710ed777 100644
--- a/tests/scenarios/cmd/sed/errors/blocked_write.yaml
+++ b/tests/scenarios/cmd/sed/errors/blocked_write.yaml
@@ -1,4 +1,5 @@
 description: The w command is blocked for safety.
+# skip: feature is intentionally blocked in rshell for security
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/sed/errors/blocked_write_cmd.yaml b/tests/scenarios/cmd/sed/errors/blocked_write_cmd.yaml
index 77967c58..2276f903 100644
--- a/tests/scenarios/cmd/sed/errors/blocked_write_cmd.yaml
+++ b/tests/scenarios/cmd/sed/errors/blocked_write_cmd.yaml
@@ -1,4 +1,5 @@
 description: The w command is blocked for safety.
+# skip: feature is intentionally blocked in rshell for security
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/sed/errors/empty_regex_addr_no_previous.yaml b/tests/scenarios/cmd/sed/errors/empty_regex_addr_no_previous.yaml
index 70e65efa..4cefbea0 100644
--- a/tests/scenarios/cmd/sed/errors/empty_regex_addr_no_previous.yaml
+++ b/tests/scenarios/cmd/sed/errors/empty_regex_addr_no_previous.yaml
@@ -2,6 +2,7 @@ description: "Empty regex address // with no previous regex produces an error."
 input:
   script: |+
     echo test | sed -n '//p'
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 expect:
   stdout: ""
diff --git a/tests/scenarios/cmd/sed/errors/invalid_backref_parse_time.yaml b/tests/scenarios/cmd/sed/errors/invalid_backref_parse_time.yaml
index 581ee279..0e5ca90e 100644
--- a/tests/scenarios/cmd/sed/errors/invalid_backref_parse_time.yaml
+++ b/tests/scenarios/cmd/sed/errors/invalid_backref_parse_time.yaml
@@ -2,6 +2,7 @@ description: "Invalid backreference in s/// rejected at parse time even when add
 input:
   script: |+
     echo x | sed '/y/s/a/\1/'
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 expect:
   stdout: ""
diff --git a/tests/scenarios/cmd/sed/errors/invalid_s_flag.yaml b/tests/scenarios/cmd/sed/errors/invalid_s_flag.yaml
index 88b37961..8c13dea5 100644
--- a/tests/scenarios/cmd/sed/errors/invalid_s_flag.yaml
+++ b/tests/scenarios/cmd/sed/errors/invalid_s_flag.yaml
@@ -7,6 +7,7 @@ input:
   allowed_paths: ["$DIR"]
   script: |+
     sed 's/a/b/z' input.txt
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 expect:
   stdout: ""
diff --git a/tests/scenarios/cmd/sed/errors/missing_file.yaml b/tests/scenarios/cmd/sed/errors/missing_file.yaml
index dace29ba..47281d52 100644
--- a/tests/scenarios/cmd/sed/errors/missing_file.yaml
+++ b/tests/scenarios/cmd/sed/errors/missing_file.yaml
@@ -3,6 +3,7 @@ setup:
   files:
     - path: dummy.txt
       content: ""
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 input:
   allowed_paths: ["$DIR"]
diff --git a/tests/scenarios/cmd/sed/errors/no_script.yaml b/tests/scenarios/cmd/sed/errors/no_script.yaml
index b3031e7d..833bc2cd 100644
--- a/tests/scenarios/cmd/sed/errors/no_script.yaml
+++ b/tests/scenarios/cmd/sed/errors/no_script.yaml
@@ -2,6 +2,7 @@ description: Sed reports an error when no script is provided.
 input:
   script: |+
     sed
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 expect:
   stdout: ""
diff --git a/tests/scenarios/cmd/sed/errors/undefined_label.yaml b/tests/scenarios/cmd/sed/errors/undefined_label.yaml
index e4aab3e3..4f26d5a0 100644
--- a/tests/scenarios/cmd/sed/errors/undefined_label.yaml
+++ b/tests/scenarios/cmd/sed/errors/undefined_label.yaml
@@ -1,4 +1,5 @@
 description: Undefined branch labels are rejected at parse time.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/sed/errors/unterminated_s.yaml b/tests/scenarios/cmd/sed/errors/unterminated_s.yaml
index d460a67e..02442551 100644
--- a/tests/scenarios/cmd/sed/errors/unterminated_s.yaml
+++ b/tests/scenarios/cmd/sed/errors/unterminated_s.yaml
@@ -7,6 +7,7 @@ input:
   allowed_paths: ["$DIR"]
   script: |+
     sed 's/a/b' input.txt
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 expect:
   stdout: ""
diff --git a/tests/scenarios/cmd/sed/errors/zero_occurrence.yaml b/tests/scenarios/cmd/sed/errors/zero_occurrence.yaml
index b3fcda68..a14c3990 100644
--- a/tests/scenarios/cmd/sed/errors/zero_occurrence.yaml
+++ b/tests/scenarios/cmd/sed/errors/zero_occurrence.yaml
@@ -1,4 +1,5 @@
 description: The s command rejects 0 as an occurrence number.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/sed/flags/help.yaml b/tests/scenarios/cmd/sed/flags/help.yaml
index d3043440..d6afec42 100644
--- a/tests/scenarios/cmd/sed/flags/help.yaml
+++ b/tests/scenarios/cmd/sed/flags/help.yaml
@@ -1,4 +1,5 @@
 description: The -h flag prints usage information and exits 0.
+# skip: rshell builtin help output differs from GNU coreutils
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/sed/pathological/no_trailing_newline.yaml b/tests/scenarios/cmd/sed/pathological/no_trailing_newline.yaml
index ee64f371..3749bfe0 100644
--- a/tests/scenarios/cmd/sed/pathological/no_trailing_newline.yaml
+++ b/tests/scenarios/cmd/sed/pathological/no_trailing_newline.yaml
@@ -12,4 +12,5 @@ expect:
   stdout: "goodbye world\n"
   stderr: ""
   exit_code: 0
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
diff --git a/tests/scenarios/cmd/sed/substitute/empty_pattern_no_previous.yaml b/tests/scenarios/cmd/sed/substitute/empty_pattern_no_previous.yaml
index 851226ce..6b9adff1 100644
--- a/tests/scenarios/cmd/sed/substitute/empty_pattern_no_previous.yaml
+++ b/tests/scenarios/cmd/sed/substitute/empty_pattern_no_previous.yaml
@@ -2,6 +2,7 @@ description: "Empty pattern in s/// with no previous regex produces an error."
 input:
   script: |+
     echo test | sed 's//X/'
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 expect:
   stdout: ""
diff --git a/tests/scenarios/cmd/sed/substitute/invalid_backref.yaml b/tests/scenarios/cmd/sed/substitute/invalid_backref.yaml
index 1478d65a..e5e964af 100644
--- a/tests/scenarios/cmd/sed/substitute/invalid_backref.yaml
+++ b/tests/scenarios/cmd/sed/substitute/invalid_backref.yaml
@@ -2,6 +2,7 @@ description: "Invalid backreference in s/// replacement is rejected."
 input:
   script: |+
     echo test | sed 's/a/\1/'
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 expect:
   stdout: ""
diff --git a/tests/scenarios/cmd/sort/errors/missing_file.yaml b/tests/scenarios/cmd/sort/errors/missing_file.yaml
index 8dc003ad..5f41cdb6 100644
--- a/tests/scenarios/cmd/sort/errors/missing_file.yaml
+++ b/tests/scenarios/cmd/sort/errors/missing_file.yaml
@@ -9,4 +9,5 @@ expect:
   stdout: ""
   stderr_contains: ["sort:"]
   exit_code: 1
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # error message format differs
diff --git a/tests/scenarios/cmd/sort/errors/unknown_flag.yaml b/tests/scenarios/cmd/sort/errors/unknown_flag.yaml
index eac9ff79..54069fee 100644
--- a/tests/scenarios/cmd/sort/errors/unknown_flag.yaml
+++ b/tests/scenarios/cmd/sort/errors/unknown_flag.yaml
@@ -11,4 +11,5 @@ expect:
   stdout: ""
   stderr_contains: ["sort:"]
   exit_code: 1
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # error message format differs
diff --git a/tests/scenarios/cmd/sort/flags/check_empty_value_reject.yaml b/tests/scenarios/cmd/sort/flags/check_empty_value_reject.yaml
index ca77086d..978d6aa0 100644
--- a/tests/scenarios/cmd/sort/flags/check_empty_value_reject.yaml
+++ b/tests/scenarios/cmd/sort/flags/check_empty_value_reject.yaml
@@ -7,6 +7,7 @@ input:
   allowed_paths: ["$DIR"]
   script: |+
     sort --check= data.txt
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 expect:
   stdout: ""
diff --git a/tests/scenarios/cmd/sort/flags/check_mixed_modes_reject.yaml b/tests/scenarios/cmd/sort/flags/check_mixed_modes_reject.yaml
index 3f8a4684..69464b62 100644
--- a/tests/scenarios/cmd/sort/flags/check_mixed_modes_reject.yaml
+++ b/tests/scenarios/cmd/sort/flags/check_mixed_modes_reject.yaml
@@ -7,6 +7,7 @@ input:
   allowed_paths: ["$DIR"]
   script: |+
     sort -c --check=silent data.txt
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 expect:
   stdout: ""
diff --git a/tests/scenarios/cmd/sort/flags/check_read_error_exit2.yaml b/tests/scenarios/cmd/sort/flags/check_read_error_exit2.yaml
index 2c647fb9..a52e10fc 100644
--- a/tests/scenarios/cmd/sort/flags/check_read_error_exit2.yaml
+++ b/tests/scenarios/cmd/sort/flags/check_read_error_exit2.yaml
@@ -3,6 +3,7 @@ input:
   allowed_paths: ["$DIR"]
   script: |+
     sort -c nonexistent.txt
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 expect:
   stdout: ""
diff --git a/tests/scenarios/cmd/sort/flags/incompatible_check_flags.yaml b/tests/scenarios/cmd/sort/flags/incompatible_check_flags.yaml
index a438071a..64167193 100644
--- a/tests/scenarios/cmd/sort/flags/incompatible_check_flags.yaml
+++ b/tests/scenarios/cmd/sort/flags/incompatible_check_flags.yaml
@@ -7,6 +7,7 @@ input:
   allowed_paths: ["$DIR"]
   script: |+
     sort -c -C data.txt
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 expect:
   stdout: ""
diff --git a/tests/scenarios/cmd/sort/flags/incompatible_flags.yaml b/tests/scenarios/cmd/sort/flags/incompatible_flags.yaml
index 75767dd9..aaf03549 100644
--- a/tests/scenarios/cmd/sort/flags/incompatible_flags.yaml
+++ b/tests/scenarios/cmd/sort/flags/incompatible_flags.yaml
@@ -7,6 +7,7 @@ input:
   allowed_paths: ["$DIR"]
   script: |+
     sort -n -d data.txt
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 expect:
   stdout: ""
diff --git a/tests/scenarios/cmd/sort/flags/incompatible_flags_per_key.yaml b/tests/scenarios/cmd/sort/flags/incompatible_flags_per_key.yaml
index 8bc558d0..436acf2d 100644
--- a/tests/scenarios/cmd/sort/flags/incompatible_flags_per_key.yaml
+++ b/tests/scenarios/cmd/sort/flags/incompatible_flags_per_key.yaml
@@ -7,6 +7,7 @@ input:
   allowed_paths: ["$DIR"]
   script: |+
     sort -k 1nd data.txt
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 expect:
   stdout: ""
diff --git a/tests/scenarios/cmd/sort/flags/key_trailing_comma.yaml b/tests/scenarios/cmd/sort/flags/key_trailing_comma.yaml
index 16cdaba5..9866e648 100644
--- a/tests/scenarios/cmd/sort/flags/key_trailing_comma.yaml
+++ b/tests/scenarios/cmd/sort/flags/key_trailing_comma.yaml
@@ -7,6 +7,7 @@ input:
   allowed_paths: ["$DIR"]
   script: |+
     sort -k 1, data.txt
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 expect:
   stdout: ""
diff --git a/tests/scenarios/cmd/sort/flags/numeric_key.yaml b/tests/scenarios/cmd/sort/flags/numeric_key.yaml
index 2c20e1b3..14f1046d 100644
--- a/tests/scenarios/cmd/sort/flags/numeric_key.yaml
+++ b/tests/scenarios/cmd/sort/flags/numeric_key.yaml
@@ -11,4 +11,5 @@ expect:
   stdout: "a:1\nb:20\nc:30\n"
   stderr: ""
   exit_code: 0
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # -k 2n glued syntax may differ
diff --git a/tests/scenarios/cmd/sort/hardening/output_flag_rejected.yaml b/tests/scenarios/cmd/sort/hardening/output_flag_rejected.yaml
index cb0f703f..5979edb6 100644
--- a/tests/scenarios/cmd/sort/hardening/output_flag_rejected.yaml
+++ b/tests/scenarios/cmd/sort/hardening/output_flag_rejected.yaml
@@ -1,5 +1,6 @@
 # Per RULES.md: -o writes to filesystem and must be rejected
 description: sort rejects the -o flag which writes to the filesystem.
+# skip: feature is intentionally blocked in rshell for security
 skip_assert_against_bash: true  # intentional restriction; -o is a valid GNU sort flag
 setup:
   files:
diff --git a/tests/scenarios/cmd/sort/hardening/outside_allowed_paths.yaml b/tests/scenarios/cmd/sort/hardening/outside_allowed_paths.yaml
index 01b6b851..cfea00ad 100644
--- a/tests/scenarios/cmd/sort/hardening/outside_allowed_paths.yaml
+++ b/tests/scenarios/cmd/sort/hardening/outside_allowed_paths.yaml
@@ -1,5 +1,6 @@
 # Per RULES.md: file access must be sandboxed via AllowedPaths
 description: sort is blocked from reading files outside the allowed paths sandbox.
+# skip: sandbox path restrictions are an rshell-specific feature
 skip_assert_against_bash: true  # intentional sandbox restriction; bash/sort can read /etc/passwd freely
 setup:
   files:
diff --git a/tests/scenarios/cmd/strings/basic/help.yaml b/tests/scenarios/cmd/strings/basic/help.yaml
index 8ca39cec..17f06657 100644
--- a/tests/scenarios/cmd/strings/basic/help.yaml
+++ b/tests/scenarios/cmd/strings/basic/help.yaml
@@ -1,4 +1,5 @@
 description: strings --help prints usage to stdout and exits 0.
+# skip: rshell builtin help output differs from GNU coreutils
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/strings/errors/empty_radix.yaml b/tests/scenarios/cmd/strings/errors/empty_radix.yaml
index f7c7557e..18270822 100644
--- a/tests/scenarios/cmd/strings/errors/empty_radix.yaml
+++ b/tests/scenarios/cmd/strings/errors/empty_radix.yaml
@@ -1,4 +1,5 @@
 description: strings --radix= with empty value prints an error and exits 1.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # GNU strings uses different error/usage format for invalid flags
 setup:
   files:
diff --git a/tests/scenarios/cmd/strings/errors/invalid_radix.yaml b/tests/scenarios/cmd/strings/errors/invalid_radix.yaml
index 675f2892..7c85c307 100644
--- a/tests/scenarios/cmd/strings/errors/invalid_radix.yaml
+++ b/tests/scenarios/cmd/strings/errors/invalid_radix.yaml
@@ -1,4 +1,5 @@
 description: strings -t with an invalid radix character prints an error and exits 1.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # GNU strings uses different error/usage format for invalid flags
 setup:
   files:
diff --git a/tests/scenarios/cmd/strings/stdin/dash_explicit.yaml b/tests/scenarios/cmd/strings/stdin/dash_explicit.yaml
index 47ee8c9c..09907585 100644
--- a/tests/scenarios/cmd/strings/stdin/dash_explicit.yaml
+++ b/tests/scenarios/cmd/strings/stdin/dash_explicit.yaml
@@ -1,4 +1,5 @@
 description: strings treats - as an alias for stdin.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # GNU strings treats - as a filename, not stdin
 input:
   script: |+
diff --git a/tests/scenarios/cmd/tail/hardening/outside_allowed_paths.yaml b/tests/scenarios/cmd/tail/hardening/outside_allowed_paths.yaml
index 76ae3bb2..ed0519fa 100644
--- a/tests/scenarios/cmd/tail/hardening/outside_allowed_paths.yaml
+++ b/tests/scenarios/cmd/tail/hardening/outside_allowed_paths.yaml
@@ -1,5 +1,6 @@
 # Per RULES.md: file access must be sandboxed via AllowedPaths
 description: tail is blocked from reading files outside the allowed paths sandbox.
+# skip: sandbox path restrictions are an rshell-specific feature
 skip_assert_against_bash: true  # intentional sandbox restriction; bash/tail can read /etc/passwd freely
 setup:
   files:
diff --git a/tests/scenarios/cmd/test/bracket/missing_bracket.yaml b/tests/scenarios/cmd/test/bracket/missing_bracket.yaml
index 0447b050..6bdfbfd9 100644
--- a/tests/scenarios/cmd/test/bracket/missing_bracket.yaml
+++ b/tests/scenarios/cmd/test/bracket/missing_bracket.yaml
@@ -1,4 +1,5 @@
 description: "[ without closing ] gives exit code 2 and error"
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # bash prefixes error with scriptname:line
 input:
   script: |+
diff --git a/tests/scenarios/cmd/test/files/directory.yaml b/tests/scenarios/cmd/test/files/directory.yaml
index 43ee34f1..28f5ce7c 100644
--- a/tests/scenarios/cmd/test/files/directory.yaml
+++ b/tests/scenarios/cmd/test/files/directory.yaml
@@ -16,4 +16,5 @@ expect:
     file: 1
   stderr: ""
   exit_code: 0
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # uses allowed_paths which the bash harness does not support
diff --git a/tests/scenarios/cmd/test/files/existence.yaml b/tests/scenarios/cmd/test/files/existence.yaml
index 40babcda..157b5b6a 100644
--- a/tests/scenarios/cmd/test/files/existence.yaml
+++ b/tests/scenarios/cmd/test/files/existence.yaml
@@ -26,4 +26,5 @@ expect:
     nofile: 1
   stderr: ""
   exit_code: 0
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # uses allowed_paths which the bash harness does not support
diff --git a/tests/scenarios/cmd/test/files/existence_a.yaml b/tests/scenarios/cmd/test/files/existence_a.yaml
index cee9d64e..0ca4b194 100644
--- a/tests/scenarios/cmd/test/files/existence_a.yaml
+++ b/tests/scenarios/cmd/test/files/existence_a.yaml
@@ -24,4 +24,5 @@ expect:
     bracket noexist: 1
   stderr: ""
   exit_code: 0
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # uses allowed_paths which the bash harness does not support
diff --git a/tests/scenarios/cmd/test/files/size.yaml b/tests/scenarios/cmd/test/files/size.yaml
index 4e10e914..175a7c9b 100644
--- a/tests/scenarios/cmd/test/files/size.yaml
+++ b/tests/scenarios/cmd/test/files/size.yaml
@@ -21,4 +21,5 @@ expect:
     missing: 1
   stderr: ""
   exit_code: 0
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # uses allowed_paths which the bash harness does not support
diff --git a/tests/scenarios/cmd/test/integers/invalid.yaml b/tests/scenarios/cmd/test/integers/invalid.yaml
index e63fe454..c4ed9d42 100644
--- a/tests/scenarios/cmd/test/integers/invalid.yaml
+++ b/tests/scenarios/cmd/test/integers/invalid.yaml
@@ -1,4 +1,5 @@
 description: "test rejects non-decimal integers with exit code 2"
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # bash prefixes error with scriptname:line
 input:
   script: |+
diff --git a/tests/scenarios/cmd/test/integers/overflow.yaml b/tests/scenarios/cmd/test/integers/overflow.yaml
index 8dc5ab46..2af3906b 100644
--- a/tests/scenarios/cmd/test/integers/overflow.yaml
+++ b/tests/scenarios/cmd/test/integers/overflow.yaml
@@ -1,4 +1,5 @@
 description: "test rejects integer overflow with exit code 2 (matches bash)"
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # bash prefixes error with scriptname:line
 input:
   script: |+
diff --git a/tests/scenarios/cmd/tr/errors/unknown_flag.yaml b/tests/scenarios/cmd/tr/errors/unknown_flag.yaml
index bfc38def..66acbaad 100644
--- a/tests/scenarios/cmd/tr/errors/unknown_flag.yaml
+++ b/tests/scenarios/cmd/tr/errors/unknown_flag.yaml
@@ -1,5 +1,6 @@
 # Per RULES.md: unknown flags must be rejected
 description: tr with unknown flag exits 1 with error.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/tr/hardening/help.yaml b/tests/scenarios/cmd/tr/hardening/help.yaml
index 21c26f41..e0867134 100644
--- a/tests/scenarios/cmd/tr/hardening/help.yaml
+++ b/tests/scenarios/cmd/tr/hardening/help.yaml
@@ -1,5 +1,6 @@
 # Per RULES.md: help flag must print usage to stdout and exit 0
 description: tr --help prints usage to stdout and exits 0.
+# skip: rshell builtin help output differs from GNU coreutils
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/uniq/errors/all_repeated_empty_method.yaml b/tests/scenarios/cmd/uniq/errors/all_repeated_empty_method.yaml
index 92b97c5c..002b5a4c 100644
--- a/tests/scenarios/cmd/uniq/errors/all_repeated_empty_method.yaml
+++ b/tests/scenarios/cmd/uniq/errors/all_repeated_empty_method.yaml
@@ -1,4 +1,5 @@
 description: uniq --all-repeated= with explicit empty method is rejected.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/uniq/errors/group_empty_method.yaml b/tests/scenarios/cmd/uniq/errors/group_empty_method.yaml
index db2cd791..0ccc6587 100644
--- a/tests/scenarios/cmd/uniq/errors/group_empty_method.yaml
+++ b/tests/scenarios/cmd/uniq/errors/group_empty_method.yaml
@@ -1,4 +1,5 @@
 description: uniq --group= with explicit empty method is rejected.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/uniq/errors/unknown_flag.yaml b/tests/scenarios/cmd/uniq/errors/unknown_flag.yaml
index 258ca647..d60a9808 100644
--- a/tests/scenarios/cmd/uniq/errors/unknown_flag.yaml
+++ b/tests/scenarios/cmd/uniq/errors/unknown_flag.yaml
@@ -1,5 +1,6 @@
 # Test unknown flag rejection
 description: uniq rejects unknown flags with exit code 1.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/cmd/unknown_cmd/common_progs/bash.yaml b/tests/scenarios/cmd/unknown_cmd/common_progs/bash.yaml
index d9f94e68..e6dbec0e 100644
--- a/tests/scenarios/cmd/unknown_cmd/common_progs/bash.yaml
+++ b/tests/scenarios/cmd/unknown_cmd/common_progs/bash.yaml
@@ -1,3 +1,4 @@
+# skip: rshell reports different error format than bash for unavailable commands
 skip_assert_against_bash: true
 description: The bash command is not a builtin and is rejected as unknown.
 input:
diff --git a/tests/scenarios/cmd/unknown_cmd/common_progs/chmod.yaml b/tests/scenarios/cmd/unknown_cmd/common_progs/chmod.yaml
index d72f136f..39061c1f 100644
--- a/tests/scenarios/cmd/unknown_cmd/common_progs/chmod.yaml
+++ b/tests/scenarios/cmd/unknown_cmd/common_progs/chmod.yaml
@@ -1,3 +1,4 @@
+# skip: rshell reports different error format than bash for unavailable commands
 skip_assert_against_bash: true
 description: The chmod command is not a builtin and is rejected as unknown.
 input:
diff --git a/tests/scenarios/cmd/unknown_cmd/common_progs/cp.yaml b/tests/scenarios/cmd/unknown_cmd/common_progs/cp.yaml
index a97d09ce..0e091d9f 100644
--- a/tests/scenarios/cmd/unknown_cmd/common_progs/cp.yaml
+++ b/tests/scenarios/cmd/unknown_cmd/common_progs/cp.yaml
@@ -1,3 +1,4 @@
+# skip: rshell reports different error format than bash for unavailable commands
 skip_assert_against_bash: true
 description: The cp command is not a builtin and is rejected as unknown.
 input:
diff --git a/tests/scenarios/cmd/unknown_cmd/common_progs/curl.yaml b/tests/scenarios/cmd/unknown_cmd/common_progs/curl.yaml
index 279be7a8..4f12e82e 100644
--- a/tests/scenarios/cmd/unknown_cmd/common_progs/curl.yaml
+++ b/tests/scenarios/cmd/unknown_cmd/common_progs/curl.yaml
@@ -1,3 +1,4 @@
+# skip: rshell reports different error format than bash for unavailable commands
 skip_assert_against_bash: true
 description: The curl command is not a builtin and is rejected as unknown.
 input:
diff --git a/tests/scenarios/cmd/unknown_cmd/common_progs/mv.yaml b/tests/scenarios/cmd/unknown_cmd/common_progs/mv.yaml
index 44a18b31..48f31825 100644
--- a/tests/scenarios/cmd/unknown_cmd/common_progs/mv.yaml
+++ b/tests/scenarios/cmd/unknown_cmd/common_progs/mv.yaml
@@ -1,3 +1,4 @@
+# skip: rshell reports different error format than bash for unavailable commands
 skip_assert_against_bash: true
 description: The mv command is not a builtin and is rejected as unknown.
 input:
diff --git a/tests/scenarios/cmd/unknown_cmd/common_progs/python.yaml b/tests/scenarios/cmd/unknown_cmd/common_progs/python.yaml
index 27a9f647..e3f386ae 100644
--- a/tests/scenarios/cmd/unknown_cmd/common_progs/python.yaml
+++ b/tests/scenarios/cmd/unknown_cmd/common_progs/python.yaml
@@ -1,3 +1,4 @@
+# skip: rshell reports different error format than bash for unavailable commands
 skip_assert_against_bash: true
 description: The python command is not a builtin and is rejected as unknown.
 input:
diff --git a/tests/scenarios/cmd/unknown_cmd/common_progs/rm.yaml b/tests/scenarios/cmd/unknown_cmd/common_progs/rm.yaml
index a65d58f2..9944a5c0 100644
--- a/tests/scenarios/cmd/unknown_cmd/common_progs/rm.yaml
+++ b/tests/scenarios/cmd/unknown_cmd/common_progs/rm.yaml
@@ -1,3 +1,4 @@
+# skip: rshell reports different error format than bash for unavailable commands
 skip_assert_against_bash: true
 description: The rm command is not a builtin and is rejected as unknown.
 input:
diff --git a/tests/scenarios/cmd/unknown_cmd/common_progs/sh.yaml b/tests/scenarios/cmd/unknown_cmd/common_progs/sh.yaml
index d9cb1edd..4762d282 100644
--- a/tests/scenarios/cmd/unknown_cmd/common_progs/sh.yaml
+++ b/tests/scenarios/cmd/unknown_cmd/common_progs/sh.yaml
@@ -1,3 +1,4 @@
+# skip: rshell reports different error format than bash for unavailable commands
 skip_assert_against_bash: true
 description: The sh command is not a builtin and is rejected as unknown.
 input:
diff --git a/tests/scenarios/cmd/unknown_cmd/common_progs/wget.yaml b/tests/scenarios/cmd/unknown_cmd/common_progs/wget.yaml
index 924a8c5c..e9c7b2bd 100644
--- a/tests/scenarios/cmd/unknown_cmd/common_progs/wget.yaml
+++ b/tests/scenarios/cmd/unknown_cmd/common_progs/wget.yaml
@@ -1,3 +1,4 @@
+# skip: rshell reports different error format than bash for unavailable commands
 skip_assert_against_bash: true
 description: The wget command is not a builtin and is rejected as unknown.
 input:
diff --git a/tests/scenarios/cmd/wc/errors/dir_single_col_width7.yaml b/tests/scenarios/cmd/wc/errors/dir_single_col_width7.yaml
index 4ed13f46..92c80a6b 100644
--- a/tests/scenarios/cmd/wc/errors/dir_single_col_width7.yaml
+++ b/tests/scenarios/cmd/wc/errors/dir_single_col_width7.yaml
@@ -1,6 +1,7 @@
 # GNU wc applies width-7 padding for non-regular files (directories) even
 # with a single column flag, when multiple files produce a total line.
 description: wc -l with directory and file uses width-7 padding.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # stderr format differs from GNU wc (PortableErr normalization)
 setup:
   files:
diff --git a/tests/scenarios/cmd/wc/errors/empty_filename.yaml b/tests/scenarios/cmd/wc/errors/empty_filename.yaml
index a00ce0ca..f14d34b5 100644
--- a/tests/scenarios/cmd/wc/errors/empty_filename.yaml
+++ b/tests/scenarios/cmd/wc/errors/empty_filename.yaml
@@ -1,4 +1,5 @@
 description: wc exits 1 and prints error for empty filename argument.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # rshell returns a clearer error than GNU wc for empty filenames
 input:
   script: |+
diff --git a/tests/scenarios/cmd/wc/errors/files0_from_rejected.yaml b/tests/scenarios/cmd/wc/errors/files0_from_rejected.yaml
index 23e3515f..5ff793cf 100644
--- a/tests/scenarios/cmd/wc/errors/files0_from_rejected.yaml
+++ b/tests/scenarios/cmd/wc/errors/files0_from_rejected.yaml
@@ -1,4 +1,5 @@
 description: wc rejects --files0-from flag (security risk).
+# skip: feature is intentionally blocked in rshell for security
 skip_assert_against_bash: true  # intentionally rejects --files0-from (GTFOBins mitigation)
 input:
   allowed_paths: ["$DIR"]
diff --git a/tests/scenarios/cmd/wc/max_line_length/fullwidth_cjk.yaml b/tests/scenarios/cmd/wc/max_line_length/fullwidth_cjk.yaml
index b8b50009..87f07805 100644
--- a/tests/scenarios/cmd/wc/max_line_length/fullwidth_cjk.yaml
+++ b/tests/scenarios/cmd/wc/max_line_length/fullwidth_cjk.yaml
@@ -1,4 +1,5 @@
 description: wc -L counts display columns, CJK characters are width 2.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # display width depends on locale; we always use Unicode width
 setup:
   files:
diff --git a/tests/scenarios/cmd/wc/max_line_length/fullwidth_emoji.yaml b/tests/scenarios/cmd/wc/max_line_length/fullwidth_emoji.yaml
index 88329252..6940a307 100644
--- a/tests/scenarios/cmd/wc/max_line_length/fullwidth_emoji.yaml
+++ b/tests/scenarios/cmd/wc/max_line_length/fullwidth_emoji.yaml
@@ -1,4 +1,5 @@
 description: wc -L counts display columns, emoji characters are width 2.
+# skip: rshell builtin error/output format differs from GNU coreutils
 skip_assert_against_bash: true  # display width depends on locale; we always use Unicode width
 setup:
   files:
diff --git a/tests/scenarios/shell/allowed_paths/cat_after_blocked_continues.yaml b/tests/scenarios/shell/allowed_paths/cat_after_blocked_continues.yaml
index c9d6119f..353e05d9 100644
--- a/tests/scenarios/shell/allowed_paths/cat_after_blocked_continues.yaml
+++ b/tests/scenarios/shell/allowed_paths/cat_after_blocked_continues.yaml
@@ -1,4 +1,5 @@
 description: "Execution continues after a blocked file access"
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/shell/allowed_paths/cat_inside_allowed.yaml b/tests/scenarios/shell/allowed_paths/cat_inside_allowed.yaml
index 7be6a841..9cea577a 100644
--- a/tests/scenarios/shell/allowed_paths/cat_inside_allowed.yaml
+++ b/tests/scenarios/shell/allowed_paths/cat_inside_allowed.yaml
@@ -1,4 +1,3 @@
-skip_assert_against_bash: true
 description: Cat can read a file inside allowed paths
 setup:
   files:
diff --git a/tests/scenarios/shell/allowed_paths/cat_outside_allowed.yaml b/tests/scenarios/shell/allowed_paths/cat_outside_allowed.yaml
index 63235f53..b08d5987 100644
--- a/tests/scenarios/shell/allowed_paths/cat_outside_allowed.yaml
+++ b/tests/scenarios/shell/allowed_paths/cat_outside_allowed.yaml
@@ -1,3 +1,4 @@
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 description: Cat cannot read a file outside allowed paths
 setup:
diff --git a/tests/scenarios/shell/allowed_paths/default_blocks_all.yaml b/tests/scenarios/shell/allowed_paths/default_blocks_all.yaml
index 3ef880d5..3ff08355 100644
--- a/tests/scenarios/shell/allowed_paths/default_blocks_all.yaml
+++ b/tests/scenarios/shell/allowed_paths/default_blocks_all.yaml
@@ -1,3 +1,4 @@
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 description: Default (no allowed_paths) blocks all file access
 setup:
diff --git a/tests/scenarios/shell/allowed_paths/dir_itself_allowed.yaml b/tests/scenarios/shell/allowed_paths/dir_itself_allowed.yaml
index 93ba9e2a..1dbd522b 100644
--- a/tests/scenarios/shell/allowed_paths/dir_itself_allowed.yaml
+++ b/tests/scenarios/shell/allowed_paths/dir_itself_allowed.yaml
@@ -1,4 +1,5 @@
 description: "When $DIR is allowed, all files under the temp directory are accessible"
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/shell/allowed_paths/dotdot_filename_allowed.yaml b/tests/scenarios/shell/allowed_paths/dotdot_filename_allowed.yaml
index fa93298f..c95a0c94 100644
--- a/tests/scenarios/shell/allowed_paths/dotdot_filename_allowed.yaml
+++ b/tests/scenarios/shell/allowed_paths/dotdot_filename_allowed.yaml
@@ -1,3 +1,4 @@
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 description: Files with names starting with .. (e.g. ..hidden) are accessible inside allowed paths
 setup:
diff --git a/tests/scenarios/shell/allowed_paths/dotdot_filename_in_subdir.yaml b/tests/scenarios/shell/allowed_paths/dotdot_filename_in_subdir.yaml
index d2074d4c..32189df6 100644
--- a/tests/scenarios/shell/allowed_paths/dotdot_filename_in_subdir.yaml
+++ b/tests/scenarios/shell/allowed_paths/dotdot_filename_in_subdir.yaml
@@ -1,3 +1,4 @@
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 description: Files with ..-prefixed names in subdirectories are accessible inside allowed paths
 setup:
diff --git a/tests/scenarios/shell/allowed_paths/empty_allowed_blocks_all.yaml b/tests/scenarios/shell/allowed_paths/empty_allowed_blocks_all.yaml
index 880021f1..18765b0e 100644
--- a/tests/scenarios/shell/allowed_paths/empty_allowed_blocks_all.yaml
+++ b/tests/scenarios/shell/allowed_paths/empty_allowed_blocks_all.yaml
@@ -1,3 +1,4 @@
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 description: Empty allowed_paths blocks all file access
 setup:
diff --git a/tests/scenarios/shell/allowed_paths/for_loop_cat_files.yaml b/tests/scenarios/shell/allowed_paths/for_loop_cat_files.yaml
index feb51686..47569556 100644
--- a/tests/scenarios/shell/allowed_paths/for_loop_cat_files.yaml
+++ b/tests/scenarios/shell/allowed_paths/for_loop_cat_files.yaml
@@ -1,4 +1,5 @@
 description: "For loop reading files from allowed path via glob"
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/shell/allowed_paths/glob_inside_allowed.yaml b/tests/scenarios/shell/allowed_paths/glob_inside_allowed.yaml
index 398d7e1f..48a13b42 100644
--- a/tests/scenarios/shell/allowed_paths/glob_inside_allowed.yaml
+++ b/tests/scenarios/shell/allowed_paths/glob_inside_allowed.yaml
@@ -1,3 +1,4 @@
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 description: Glob expansion works inside allowed paths
 setup:
diff --git a/tests/scenarios/shell/allowed_paths/glob_no_match_in_allowed.yaml b/tests/scenarios/shell/allowed_paths/glob_no_match_in_allowed.yaml
index ba98c80b..9e314485 100644
--- a/tests/scenarios/shell/allowed_paths/glob_no_match_in_allowed.yaml
+++ b/tests/scenarios/shell/allowed_paths/glob_no_match_in_allowed.yaml
@@ -1,4 +1,5 @@
 description: "Glob pattern with no matches inside allowed path returns literal pattern"
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/shell/allowed_paths/glob_outside_allowed.yaml b/tests/scenarios/shell/allowed_paths/glob_outside_allowed.yaml
index 257b2bf3..3439e5d1 100644
--- a/tests/scenarios/shell/allowed_paths/glob_outside_allowed.yaml
+++ b/tests/scenarios/shell/allowed_paths/glob_outside_allowed.yaml
@@ -1,3 +1,4 @@
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 description: Glob expansion on directory outside allowed paths returns literal pattern
 setup:
diff --git a/tests/scenarios/shell/allowed_paths/heredoc_unaffected.yaml b/tests/scenarios/shell/allowed_paths/heredoc_unaffected.yaml
index bf58ebcc..3650a0cc 100644
--- a/tests/scenarios/shell/allowed_paths/heredoc_unaffected.yaml
+++ b/tests/scenarios/shell/allowed_paths/heredoc_unaffected.yaml
@@ -1,4 +1,5 @@
 description: "Heredocs work regardless of AllowedPaths since they do not access the filesystem"
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 input:
   script: |
diff --git a/tests/scenarios/shell/allowed_paths/multiple_allowed_dirs.yaml b/tests/scenarios/shell/allowed_paths/multiple_allowed_dirs.yaml
index d014bdfe..cc6544e8 100644
--- a/tests/scenarios/shell/allowed_paths/multiple_allowed_dirs.yaml
+++ b/tests/scenarios/shell/allowed_paths/multiple_allowed_dirs.yaml
@@ -1,4 +1,5 @@
 description: "Multiple allowed directories grant access to files in each one"
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/shell/allowed_paths/multiple_allowed_one_blocked.yaml b/tests/scenarios/shell/allowed_paths/multiple_allowed_one_blocked.yaml
index 5f109057..34019cb4 100644
--- a/tests/scenarios/shell/allowed_paths/multiple_allowed_one_blocked.yaml
+++ b/tests/scenarios/shell/allowed_paths/multiple_allowed_one_blocked.yaml
@@ -1,4 +1,5 @@
 description: "Multiple allowed directories still block access outside them"
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/shell/allowed_paths/multiple_cat_same_dir.yaml b/tests/scenarios/shell/allowed_paths/multiple_cat_same_dir.yaml
index 35aaf5a6..830a1ae3 100644
--- a/tests/scenarios/shell/allowed_paths/multiple_cat_same_dir.yaml
+++ b/tests/scenarios/shell/allowed_paths/multiple_cat_same_dir.yaml
@@ -1,4 +1,5 @@
 description: "Multiple cat commands reading different files from the same allowed directory"
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/shell/allowed_paths/nonexistent_file_in_allowed.yaml b/tests/scenarios/shell/allowed_paths/nonexistent_file_in_allowed.yaml
index c44dc99a..b2fa8a36 100644
--- a/tests/scenarios/shell/allowed_paths/nonexistent_file_in_allowed.yaml
+++ b/tests/scenarios/shell/allowed_paths/nonexistent_file_in_allowed.yaml
@@ -1,4 +1,5 @@
 description: "Non-existent file inside allowed directory returns an error"
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/shell/allowed_paths/pipe_inside_allowed.yaml b/tests/scenarios/shell/allowed_paths/pipe_inside_allowed.yaml
index 67984e19..fc21e8ec 100644
--- a/tests/scenarios/shell/allowed_paths/pipe_inside_allowed.yaml
+++ b/tests/scenarios/shell/allowed_paths/pipe_inside_allowed.yaml
@@ -1,4 +1,5 @@
 description: "Piping cat output from a file in allowed path works"
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/shell/allowed_paths/redirect_from_one_cat_from_other.yaml b/tests/scenarios/shell/allowed_paths/redirect_from_one_cat_from_other.yaml
index 0dd5e866..a9d51277 100644
--- a/tests/scenarios/shell/allowed_paths/redirect_from_one_cat_from_other.yaml
+++ b/tests/scenarios/shell/allowed_paths/redirect_from_one_cat_from_other.yaml
@@ -1,4 +1,5 @@
 description: "Redirect input from one allowed path and cat from another both work"
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/shell/allowed_paths/redirect_inside_allowed.yaml b/tests/scenarios/shell/allowed_paths/redirect_inside_allowed.yaml
index 4f82c615..73106c48 100644
--- a/tests/scenarios/shell/allowed_paths/redirect_inside_allowed.yaml
+++ b/tests/scenarios/shell/allowed_paths/redirect_inside_allowed.yaml
@@ -1,3 +1,4 @@
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 description: Input redirect can read file inside allowed paths
 setup:
diff --git a/tests/scenarios/shell/allowed_paths/redirect_outside_allowed.yaml b/tests/scenarios/shell/allowed_paths/redirect_outside_allowed.yaml
index a74bdef9..b23e2fd3 100644
--- a/tests/scenarios/shell/allowed_paths/redirect_outside_allowed.yaml
+++ b/tests/scenarios/shell/allowed_paths/redirect_outside_allowed.yaml
@@ -1,3 +1,4 @@
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 description: Input redirect cannot read file outside allowed paths
 setup:
diff --git a/tests/scenarios/shell/allowed_paths/redirect_variable_inside.yaml b/tests/scenarios/shell/allowed_paths/redirect_variable_inside.yaml
index ec17c1a2..89d088d6 100644
--- a/tests/scenarios/shell/allowed_paths/redirect_variable_inside.yaml
+++ b/tests/scenarios/shell/allowed_paths/redirect_variable_inside.yaml
@@ -1,4 +1,5 @@
 description: "Input redirect using a variable-expanded path inside allowed path"
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/shell/allowed_paths/redirect_variable_outside.yaml b/tests/scenarios/shell/allowed_paths/redirect_variable_outside.yaml
index ab3b2dc3..d17efa19 100644
--- a/tests/scenarios/shell/allowed_paths/redirect_variable_outside.yaml
+++ b/tests/scenarios/shell/allowed_paths/redirect_variable_outside.yaml
@@ -1,4 +1,5 @@
 description: "Input redirect using a variable-expanded path outside allowed path is blocked"
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/shell/allowed_paths/subdir_of_allowed.yaml b/tests/scenarios/shell/allowed_paths/subdir_of_allowed.yaml
index 65679b6c..7c1376ca 100644
--- a/tests/scenarios/shell/allowed_paths/subdir_of_allowed.yaml
+++ b/tests/scenarios/shell/allowed_paths/subdir_of_allowed.yaml
@@ -1,4 +1,5 @@
 description: "Files in nested subdirectories inside an allowed path are accessible"
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/shell/allowed_paths/symlink_chain_escape.yaml b/tests/scenarios/shell/allowed_paths/symlink_chain_escape.yaml
index 0e8039f6..dca6aaa5 100644
--- a/tests/scenarios/shell/allowed_paths/symlink_chain_escape.yaml
+++ b/tests/scenarios/shell/allowed_paths/symlink_chain_escape.yaml
@@ -1,3 +1,4 @@
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 description: Chained symlinks that escape allowed dir are blocked by os.Root
 setup:
diff --git a/tests/scenarios/shell/allowed_paths/symlink_escape_to_dir.yaml b/tests/scenarios/shell/allowed_paths/symlink_escape_to_dir.yaml
index 89b37aa3..183659c8 100644
--- a/tests/scenarios/shell/allowed_paths/symlink_escape_to_dir.yaml
+++ b/tests/scenarios/shell/allowed_paths/symlink_escape_to_dir.yaml
@@ -1,3 +1,4 @@
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 description: Symlink inside allowed dir pointing to directory outside is blocked by os.Root
 setup:
diff --git a/tests/scenarios/shell/allowed_paths/symlink_escape_to_file.yaml b/tests/scenarios/shell/allowed_paths/symlink_escape_to_file.yaml
index 240e9d8a..2aa29d60 100644
--- a/tests/scenarios/shell/allowed_paths/symlink_escape_to_file.yaml
+++ b/tests/scenarios/shell/allowed_paths/symlink_escape_to_file.yaml
@@ -1,3 +1,4 @@
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 description: Symlink inside allowed dir pointing to file outside is blocked by os.Root
 setup:
diff --git a/tests/scenarios/shell/allowed_paths/symlink_redirect_escape.yaml b/tests/scenarios/shell/allowed_paths/symlink_redirect_escape.yaml
index c58189ac..892cb764 100644
--- a/tests/scenarios/shell/allowed_paths/symlink_redirect_escape.yaml
+++ b/tests/scenarios/shell/allowed_paths/symlink_redirect_escape.yaml
@@ -1,3 +1,4 @@
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 description: Redirect via symlink that escapes allowed dir is blocked by os.Root
 setup:
diff --git a/tests/scenarios/shell/allowed_paths/symlink_within_allowed.yaml b/tests/scenarios/shell/allowed_paths/symlink_within_allowed.yaml
index 8f1475bc..db1a1f03 100644
--- a/tests/scenarios/shell/allowed_paths/symlink_within_allowed.yaml
+++ b/tests/scenarios/shell/allowed_paths/symlink_within_allowed.yaml
@@ -1,3 +1,4 @@
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 description: Symlink pointing to another file inside the same allowed dir works
 setup:
diff --git a/tests/scenarios/shell/allowed_paths/traversal_blocked.yaml b/tests/scenarios/shell/allowed_paths/traversal_blocked.yaml
index 5100927c..c9429359 100644
--- a/tests/scenarios/shell/allowed_paths/traversal_blocked.yaml
+++ b/tests/scenarios/shell/allowed_paths/traversal_blocked.yaml
@@ -1,3 +1,4 @@
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 description: Path traversal with .. is blocked by os.Root
 setup:
diff --git a/tests/scenarios/shell/allowed_paths/traversal_to_sibling.yaml b/tests/scenarios/shell/allowed_paths/traversal_to_sibling.yaml
index 260e5dd5..c7121014 100644
--- a/tests/scenarios/shell/allowed_paths/traversal_to_sibling.yaml
+++ b/tests/scenarios/shell/allowed_paths/traversal_to_sibling.yaml
@@ -1,4 +1,5 @@
 description: "Path traversal from allowed dir to sibling dir is blocked"
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/shell/allowed_paths/variable_path_inside.yaml b/tests/scenarios/shell/allowed_paths/variable_path_inside.yaml
index b4bd68f6..5ce41c4a 100644
--- a/tests/scenarios/shell/allowed_paths/variable_path_inside.yaml
+++ b/tests/scenarios/shell/allowed_paths/variable_path_inside.yaml
@@ -1,4 +1,5 @@
 description: "Variable-expanded file path inside allowed path is accessible"
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/shell/allowed_paths/variable_path_outside.yaml b/tests/scenarios/shell/allowed_paths/variable_path_outside.yaml
index f235a60d..26a06a0f 100644
--- a/tests/scenarios/shell/allowed_paths/variable_path_outside.yaml
+++ b/tests/scenarios/shell/allowed_paths/variable_path_outside.yaml
@@ -1,4 +1,5 @@
 description: "Variable-expanded file path outside allowed path is blocked"
+# skip: allowed_paths sandbox restriction is an rshell-specific feature not present in bash
 skip_assert_against_bash: true
 setup:
   files:
diff --git a/tests/scenarios/shell/blocked_commands/arithmetic_cmd.yaml b/tests/scenarios/shell/blocked_commands/arithmetic_cmd.yaml
index eb26d094..84724e7a 100644
--- a/tests/scenarios/shell/blocked_commands/arithmetic_cmd.yaml
+++ b/tests/scenarios/shell/blocked_commands/arithmetic_cmd.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Arithmetic commands are not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_commands/background.yaml b/tests/scenarios/shell/blocked_commands/background.yaml
index ba49257a..96d8acb3 100644
--- a/tests/scenarios/shell/blocked_commands/background.yaml
+++ b/tests/scenarios/shell/blocked_commands/background.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Background execution (&) is blocked.
 input:
diff --git a/tests/scenarios/shell/blocked_commands/blocked_after_valid.yaml b/tests/scenarios/shell/blocked_commands/blocked_after_valid.yaml
index f5453ac0..86c810df 100644
--- a/tests/scenarios/shell/blocked_commands/blocked_after_valid.yaml
+++ b/tests/scenarios/shell/blocked_commands/blocked_after_valid.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: A blocked command after valid commands still causes rejection.
 input:
diff --git a/tests/scenarios/shell/blocked_commands/blocked_command_after_valid.yaml b/tests/scenarios/shell/blocked_commands/blocked_command_after_valid.yaml
index af75c828..4ceffae7 100644
--- a/tests/scenarios/shell/blocked_commands/blocked_command_after_valid.yaml
+++ b/tests/scenarios/shell/blocked_commands/blocked_command_after_valid.yaml
@@ -7,6 +7,6 @@ input:
     select x in a b; do echo "$x"; done
 expect:
   stdout: ""
-  stderr_contains:
-    - "not supported"
+  stderr: |+
+    select statements are not supported
   exit_code: 2
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/alias.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/alias.yaml
new file mode 100644
index 00000000..e7067c37
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/alias.yaml
@@ -0,0 +1,11 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Alias builtin is intentionally blocked in the restricted shell.
+input:
+  script: |+
+    alias ll='ls -l'
+expect:
+  stdout: ""
+  stderr: |+
+    alias: command not found
+  exit_code: 127
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/arithmetic_expansion.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/arithmetic_expansion.yaml
new file mode 100644
index 00000000..fd6f08aa
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/arithmetic_expansion.yaml
@@ -0,0 +1,11 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Arithmetic expansion is intentionally blocked in the restricted shell.
+input:
+  script: |+
+    echo $((1+2))
+expect:
+  stdout: ""
+  stderr: |+
+    arithmetic expansion is not supported
+  exit_code: 2
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/backtick_substitution.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/backtick_substitution.yaml
new file mode 100644
index 00000000..eb24a65c
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/backtick_substitution.yaml
@@ -0,0 +1,11 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Backtick command substitution is intentionally blocked in the restricted shell.
+input:
+  script: |+
+    echo `echo hello`
+expect:
+  stdout: ""
+  stderr: |+
+    command substitution is not supported
+  exit_code: 2
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/cd.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/cd.yaml
new file mode 100644
index 00000000..bfb116b8
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/cd.yaml
@@ -0,0 +1,11 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Cd builtin is intentionally blocked in the restricted shell.
+input:
+  script: |+
+    cd /tmp
+expect:
+  stdout: ""
+  stderr: |+
+    cd: command not found
+  exit_code: 127
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/command.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/command.yaml
new file mode 100644
index 00000000..c95fb5ea
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/command.yaml
@@ -0,0 +1,11 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Command builtin is intentionally blocked in the restricted shell.
+input:
+  script: |+
+    command echo hello
+expect:
+  stdout: ""
+  stderr: |+
+    command: command not found
+  exit_code: 127
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/command_substitution.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/command_substitution.yaml
new file mode 100644
index 00000000..0f398e5b
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/command_substitution.yaml
@@ -0,0 +1,11 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Command substitution is intentionally blocked in the restricted shell.
+input:
+  script: |+
+    echo $(echo hello)
+expect:
+  stdout: ""
+  stderr: |+
+    command substitution is not supported
+  exit_code: 2
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/dot_source.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/dot_source.yaml
new file mode 100644
index 00000000..d5fc1783
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/dot_source.yaml
@@ -0,0 +1,11 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Dot/source builtin is intentionally blocked in the restricted shell.
+input:
+  script: |+
+    . /dev/null
+expect:
+  stdout: ""
+  stderr: |+
+    .: command not found
+  exit_code: 127
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/exec.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/exec.yaml
new file mode 100644
index 00000000..e50bc024
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/exec.yaml
@@ -0,0 +1,11 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Exec builtin is intentionally blocked in the restricted shell.
+input:
+  script: |+
+    exec echo hello
+expect:
+  stdout: ""
+  stderr: |+
+    exec: command not found
+  exit_code: 127
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/getopts.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/getopts.yaml
new file mode 100644
index 00000000..a21e7878
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/getopts.yaml
@@ -0,0 +1,11 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Getopts builtin is intentionally blocked in the restricted shell.
+input:
+  script: |+
+    getopts abc opt
+expect:
+  stdout: ""
+  stderr: |+
+    getopts: command not found
+  exit_code: 127
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/param_alternative.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/param_alternative.yaml
new file mode 100644
index 00000000..628a11d3
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/param_alternative.yaml
@@ -0,0 +1,12 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Parameter expansion alternative operation is intentionally blocked.
+input:
+  script: |+
+    a=foo
+    echo ${a+bar}
+expect:
+  stdout: ""
+  stderr: |+
+    ${var} operations (defaults, pattern removal, case conversion) are not supported
+  exit_code: 2
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/param_assign_default.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/param_assign_default.yaml
new file mode 100644
index 00000000..9a086da0
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/param_assign_default.yaml
@@ -0,0 +1,11 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Parameter expansion assign-default operation is intentionally blocked.
+input:
+  script: |+
+    echo ${a=default}
+expect:
+  stdout: ""
+  stderr: |+
+    ${var} operations (defaults, pattern removal, case conversion) are not supported
+  exit_code: 2
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/param_default_value.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/param_default_value.yaml
new file mode 100644
index 00000000..98cd930c
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/param_default_value.yaml
@@ -0,0 +1,11 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Parameter expansion operations are intentionally blocked.
+input:
+  script: |+
+    echo ${a-default}
+expect:
+  stdout: ""
+  stderr: |+
+    ${var} operations (defaults, pattern removal, case conversion) are not supported
+  exit_code: 2
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/param_error_unset.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/param_error_unset.yaml
new file mode 100644
index 00000000..490b5292
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/param_error_unset.yaml
@@ -0,0 +1,11 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Parameter expansion error operation is intentionally blocked.
+input:
+  script: |+
+    echo ${a?error message}
+expect:
+  stdout: ""
+  stderr: |+
+    ${var} operations (defaults, pattern removal, case conversion) are not supported
+  exit_code: 2
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/param_prefix_removal.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/param_prefix_removal.yaml
new file mode 100644
index 00000000..d15514e5
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/param_prefix_removal.yaml
@@ -0,0 +1,12 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Prefix removal parameter expansion is intentionally blocked.
+input:
+  script: |+
+    a=foobar
+    echo ${a#foo}
+expect:
+  stdout: ""
+  stderr: |+
+    ${var} operations (defaults, pattern removal, case conversion) are not supported
+  exit_code: 2
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/param_string_length.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/param_string_length.yaml
new file mode 100644
index 00000000..392eb67d
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/param_string_length.yaml
@@ -0,0 +1,12 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: String length parameter expansion is intentionally blocked.
+input:
+  script: |+
+    a=hello
+    echo ${#a}
+expect:
+  stdout: ""
+  stderr: |+
+    ${#var} is not supported
+  exit_code: 2
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/param_suffix_removal.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/param_suffix_removal.yaml
new file mode 100644
index 00000000..5973db32
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/param_suffix_removal.yaml
@@ -0,0 +1,12 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Suffix removal parameter expansion is intentionally blocked.
+input:
+  script: |+
+    a=foobar
+    echo ${a%bar}
+expect:
+  stdout: ""
+  stderr: |+
+    ${var} operations (defaults, pattern removal, case conversion) are not supported
+  exit_code: 2
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/positional_params.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/positional_params.yaml
new file mode 100644
index 00000000..e62b0b0f
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/positional_params.yaml
@@ -0,0 +1,11 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Positional parameters are intentionally blocked.
+input:
+  script: |+
+    echo $1
+expect:
+  stdout: ""
+  stderr: |+
+    $1 is not supported
+  exit_code: 2
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/read.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/read.yaml
new file mode 100644
index 00000000..fd78865b
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/read.yaml
@@ -0,0 +1,11 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Read builtin is intentionally blocked in the restricted shell.
+input:
+  script: |+
+    echo hello | read var
+expect:
+  stdout: ""
+  stderr: |+
+    read: command not found
+  exit_code: 127
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/return.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/return.yaml
new file mode 100644
index 00000000..8f09a25e
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/return.yaml
@@ -0,0 +1,11 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Return builtin is intentionally blocked in the restricted shell.
+input:
+  script: |+
+    return 0
+expect:
+  stdout: ""
+  stderr: |+
+    return: command not found
+  exit_code: 127
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/set.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/set.yaml
new file mode 100644
index 00000000..2bb68429
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/set.yaml
@@ -0,0 +1,11 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Set builtin is intentionally blocked in the restricted shell.
+input:
+  script: |+
+    set -- a b c
+expect:
+  stdout: ""
+  stderr: |+
+    set: command not found
+  exit_code: 127
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/shift.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/shift.yaml
new file mode 100644
index 00000000..14db0661
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/shift.yaml
@@ -0,0 +1,11 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Shift builtin is intentionally blocked in the restricted shell.
+input:
+  script: |+
+    shift
+expect:
+  stdout: ""
+  stderr: |+
+    shift: command not found
+  exit_code: 127
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/tilde_expansion.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/tilde_expansion.yaml
new file mode 100644
index 00000000..f4dcdd8f
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/tilde_expansion.yaml
@@ -0,0 +1,11 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Tilde expansion is intentionally not supported in the restricted shell.
+input:
+  script: |+
+    echo ~
+expect:
+  stdout: ""
+  stderr: |+
+    tilde expansion is not supported
+  exit_code: 2
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/trap.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/trap.yaml
new file mode 100644
index 00000000..baa2c2c8
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/trap.yaml
@@ -0,0 +1,11 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Trap builtin is intentionally blocked in the restricted shell.
+input:
+  script: |+
+    trap 'echo trapped' INT
+expect:
+  stdout: ""
+  stderr: |+
+    trap: command not found
+  exit_code: 127
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/umask.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/umask.yaml
new file mode 100644
index 00000000..6f46bc04
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/umask.yaml
@@ -0,0 +1,11 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Umask builtin is intentionally blocked in the restricted shell.
+input:
+  script: |+
+    umask
+expect:
+  stdout: ""
+  stderr: |+
+    umask: command not found
+  exit_code: 127
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/unset.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/unset.yaml
new file mode 100644
index 00000000..5cff318c
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/unset.yaml
@@ -0,0 +1,12 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Unset builtin is intentionally blocked in the restricted shell.
+input:
+  script: |+
+    a=hello
+    unset a
+expect:
+  stdout: ""
+  stderr: |+
+    unset: command not found
+  exit_code: 127
diff --git a/tests/scenarios/shell/blocked_commands/builtins_and_features/wait.yaml b/tests/scenarios/shell/blocked_commands/builtins_and_features/wait.yaml
new file mode 100644
index 00000000..d2e83238
--- /dev/null
+++ b/tests/scenarios/shell/blocked_commands/builtins_and_features/wait.yaml
@@ -0,0 +1,11 @@
+# skip: feature is intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Wait builtin is intentionally blocked in the restricted shell.
+input:
+  script: |+
+    wait
+expect:
+  stdout: ""
+  stderr: |+
+    wait: command not found
+  exit_code: 127
diff --git a/tests/scenarios/shell/blocked_commands/c_style_for.yaml b/tests/scenarios/shell/blocked_commands/c_style_for.yaml
index 71d59191..8b934fa0 100644
--- a/tests/scenarios/shell/blocked_commands/c_style_for.yaml
+++ b/tests/scenarios/shell/blocked_commands/c_style_for.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: C-style for loops are not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_commands/case_statement.yaml b/tests/scenarios/shell/blocked_commands/case_statement.yaml
index 6397c46f..d80d0b7f 100644
--- a/tests/scenarios/shell/blocked_commands/case_statement.yaml
+++ b/tests/scenarios/shell/blocked_commands/case_statement.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Case statements are not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_commands/coproc.yaml b/tests/scenarios/shell/blocked_commands/coproc.yaml
index 765986f0..9e18da59 100644
--- a/tests/scenarios/shell/blocked_commands/coproc.yaml
+++ b/tests/scenarios/shell/blocked_commands/coproc.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Coprocesses are not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_commands/declare.yaml b/tests/scenarios/shell/blocked_commands/declare.yaml
index 7863f81c..ac38c1c2 100644
--- a/tests/scenarios/shell/blocked_commands/declare.yaml
+++ b/tests/scenarios/shell/blocked_commands/declare.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: declare is not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_commands/export.yaml b/tests/scenarios/shell/blocked_commands/export.yaml
index ecc4bc56..f9bf624c 100644
--- a/tests/scenarios/shell/blocked_commands/export.yaml
+++ b/tests/scenarios/shell/blocked_commands/export.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: export is not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_commands/extglob.yaml b/tests/scenarios/shell/blocked_commands/extglob.yaml
index bf7bc1a0..ea534bd3 100644
--- a/tests/scenarios/shell/blocked_commands/extglob.yaml
+++ b/tests/scenarios/shell/blocked_commands/extglob.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Extended globbing is not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_commands/function_decl.yaml b/tests/scenarios/shell/blocked_commands/function_decl.yaml
index b764986d..21a087f1 100644
--- a/tests/scenarios/shell/blocked_commands/function_decl.yaml
+++ b/tests/scenarios/shell/blocked_commands/function_decl.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Function declarations are not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_commands/let.yaml b/tests/scenarios/shell/blocked_commands/let.yaml
index 57c4db46..f3472274 100644
--- a/tests/scenarios/shell/blocked_commands/let.yaml
+++ b/tests/scenarios/shell/blocked_commands/let.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: let is not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_commands/local.yaml b/tests/scenarios/shell/blocked_commands/local.yaml
index 35298e7b..665915f5 100644
--- a/tests/scenarios/shell/blocked_commands/local.yaml
+++ b/tests/scenarios/shell/blocked_commands/local.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: local is not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_commands/pipe_all.yaml b/tests/scenarios/shell/blocked_commands/pipe_all.yaml
index e600587b..b7c50c49 100644
--- a/tests/scenarios/shell/blocked_commands/pipe_all.yaml
+++ b/tests/scenarios/shell/blocked_commands/pipe_all.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Pipe all (|&) is not supported (bash extension).
 input:
diff --git a/tests/scenarios/shell/blocked_commands/process_sub.yaml b/tests/scenarios/shell/blocked_commands/process_sub.yaml
index 25cc0345..9dea3a08 100644
--- a/tests/scenarios/shell/blocked_commands/process_sub.yaml
+++ b/tests/scenarios/shell/blocked_commands/process_sub.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Process substitution is not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_commands/readonly.yaml b/tests/scenarios/shell/blocked_commands/readonly.yaml
index 96be0bce..bf70edb6 100644
--- a/tests/scenarios/shell/blocked_commands/readonly.yaml
+++ b/tests/scenarios/shell/blocked_commands/readonly.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: readonly is not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_commands/select_clause.yaml b/tests/scenarios/shell/blocked_commands/select_clause.yaml
index 8bac5a82..6276e58d 100644
--- a/tests/scenarios/shell/blocked_commands/select_clause.yaml
+++ b/tests/scenarios/shell/blocked_commands/select_clause.yaml
@@ -6,6 +6,6 @@ input:
     select x in a b c; do echo $x; break; done
 expect:
   stdout: ""
-  stderr_contains:
-    - "not supported"
+  stderr: |+
+    select statements are not supported
   exit_code: 2
diff --git a/tests/scenarios/shell/blocked_commands/select_statement.yaml b/tests/scenarios/shell/blocked_commands/select_statement.yaml
index 7ac0c11c..1b869a5b 100644
--- a/tests/scenarios/shell/blocked_commands/select_statement.yaml
+++ b/tests/scenarios/shell/blocked_commands/select_statement.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Select statements are not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_commands/subshell.yaml b/tests/scenarios/shell/blocked_commands/subshell.yaml
index 4189ae94..81bdffe7 100644
--- a/tests/scenarios/shell/blocked_commands/subshell.yaml
+++ b/tests/scenarios/shell/blocked_commands/subshell.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Subshells are not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_commands/test_clause.yaml b/tests/scenarios/shell/blocked_commands/test_clause.yaml
index 76fd47d1..a74219ac 100644
--- a/tests/scenarios/shell/blocked_commands/test_clause.yaml
+++ b/tests/scenarios/shell/blocked_commands/test_clause.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Test expressions are not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_commands/time.yaml b/tests/scenarios/shell/blocked_commands/time.yaml
index a51cd9c8..7722cd92 100644
--- a/tests/scenarios/shell/blocked_commands/time.yaml
+++ b/tests/scenarios/shell/blocked_commands/time.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: time is not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_commands/until_loop.yaml b/tests/scenarios/shell/blocked_commands/until_loop.yaml
index 529917c8..0e25c133 100644
--- a/tests/scenarios/shell/blocked_commands/until_loop.yaml
+++ b/tests/scenarios/shell/blocked_commands/until_loop.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Until loops are not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_commands/while_loop.yaml b/tests/scenarios/shell/blocked_commands/while_loop.yaml
index 5c7a3501..13b0d2c9 100644
--- a/tests/scenarios/shell/blocked_commands/while_loop.yaml
+++ b/tests/scenarios/shell/blocked_commands/while_loop.yaml
@@ -1,3 +1,4 @@
+# skip: feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: While loops are not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_redirects/append_all.yaml b/tests/scenarios/shell/blocked_redirects/append_all.yaml
index 1ebe4fca..a5cf0355 100644
--- a/tests/scenarios/shell/blocked_redirects/append_all.yaml
+++ b/tests/scenarios/shell/blocked_redirects/append_all.yaml
@@ -1,3 +1,4 @@
+# skip: redirect type is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Append all (&>>) is not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_redirects/blocked_after_valid.yaml b/tests/scenarios/shell/blocked_redirects/blocked_after_valid.yaml
index 820b7c2f..6b8805e9 100644
--- a/tests/scenarios/shell/blocked_redirects/blocked_after_valid.yaml
+++ b/tests/scenarios/shell/blocked_redirects/blocked_after_valid.yaml
@@ -1,3 +1,4 @@
+# skip: redirect type is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Write redirect after valid commands still causes rejection.
 input:
diff --git a/tests/scenarios/shell/blocked_redirects/dup_close_fd.yaml b/tests/scenarios/shell/blocked_redirects/dup_close_fd.yaml
index 3e0a0095..ebce64c5 100644
--- a/tests/scenarios/shell/blocked_redirects/dup_close_fd.yaml
+++ b/tests/scenarios/shell/blocked_redirects/dup_close_fd.yaml
@@ -1,3 +1,4 @@
+# skip: redirect type is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Close-fd syntax (>&-) is blocked.
 input:
diff --git a/tests/scenarios/shell/blocked_redirects/dup_in.yaml b/tests/scenarios/shell/blocked_redirects/dup_in.yaml
index a645eeda..a0427bd6 100644
--- a/tests/scenarios/shell/blocked_redirects/dup_in.yaml
+++ b/tests/scenarios/shell/blocked_redirects/dup_in.yaml
@@ -1,3 +1,4 @@
+# skip: redirect type is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Input fd duplication (<&N) is blocked.
 input:
diff --git a/tests/scenarios/shell/blocked_redirects/dup_out.yaml b/tests/scenarios/shell/blocked_redirects/dup_out.yaml
index 9a3021db..ade1d44f 100644
--- a/tests/scenarios/shell/blocked_redirects/dup_out.yaml
+++ b/tests/scenarios/shell/blocked_redirects/dup_out.yaml
@@ -1,3 +1,4 @@
+# skip: redirect type is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Output fd duplication (>&N) with unsupported fd is blocked.
 input:
diff --git a/tests/scenarios/shell/blocked_redirects/herestring.yaml b/tests/scenarios/shell/blocked_redirects/herestring.yaml
index 1e9ae8e0..31abf02e 100644
--- a/tests/scenarios/shell/blocked_redirects/herestring.yaml
+++ b/tests/scenarios/shell/blocked_redirects/herestring.yaml
@@ -1,3 +1,4 @@
+# skip: redirect type is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Herestring (<<<) is not supported (bash extension).
 input:
diff --git a/tests/scenarios/shell/blocked_redirects/input_fd3_blocked.yaml b/tests/scenarios/shell/blocked_redirects/input_fd3_blocked.yaml
index 0f738b2c..b305d735 100644
--- a/tests/scenarios/shell/blocked_redirects/input_fd3_blocked.yaml
+++ b/tests/scenarios/shell/blocked_redirects/input_fd3_blocked.yaml
@@ -1,3 +1,4 @@
+# skip: redirect type is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Unsupported input fd (3< file) is rejected at validation time.
 input:
diff --git a/tests/scenarios/shell/blocked_redirects/read_write.yaml b/tests/scenarios/shell/blocked_redirects/read_write.yaml
index 62ef919f..7b73f28d 100644
--- a/tests/scenarios/shell/blocked_redirects/read_write.yaml
+++ b/tests/scenarios/shell/blocked_redirects/read_write.yaml
@@ -1,3 +1,4 @@
+# skip: redirect type is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Read-write redirection (<>) is not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_redirects/stderr_write.yaml b/tests/scenarios/shell/blocked_redirects/stderr_write.yaml
index 07d6399c..70afe4c5 100644
--- a/tests/scenarios/shell/blocked_redirects/stderr_write.yaml
+++ b/tests/scenarios/shell/blocked_redirects/stderr_write.yaml
@@ -1,3 +1,4 @@
+# skip: redirect type is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Stderr redirection (2>) is not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_redirects/variable_redirect_target.yaml b/tests/scenarios/shell/blocked_redirects/variable_redirect_target.yaml
index 61526f10..d6f38c76 100644
--- a/tests/scenarios/shell/blocked_redirects/variable_redirect_target.yaml
+++ b/tests/scenarios/shell/blocked_redirects/variable_redirect_target.yaml
@@ -1,3 +1,4 @@
+# skip: redirect type is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Variable expansion in redirect target is blocked (even if it resolves to /dev/null).
 input:
diff --git a/tests/scenarios/shell/blocked_redirects/write_all.yaml b/tests/scenarios/shell/blocked_redirects/write_all.yaml
index dd9fc3b2..5b52ec70 100644
--- a/tests/scenarios/shell/blocked_redirects/write_all.yaml
+++ b/tests/scenarios/shell/blocked_redirects/write_all.yaml
@@ -1,3 +1,4 @@
+# skip: redirect type is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Redirect all (&>) is not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_redirects/write_append.yaml b/tests/scenarios/shell/blocked_redirects/write_append.yaml
index e50b2137..97adc6d9 100644
--- a/tests/scenarios/shell/blocked_redirects/write_append.yaml
+++ b/tests/scenarios/shell/blocked_redirects/write_append.yaml
@@ -1,3 +1,4 @@
+# skip: redirect type is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Append redirection (>>) is not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_redirects/write_clobber.yaml b/tests/scenarios/shell/blocked_redirects/write_clobber.yaml
index dd60fbe6..60bc1f02 100644
--- a/tests/scenarios/shell/blocked_redirects/write_clobber.yaml
+++ b/tests/scenarios/shell/blocked_redirects/write_clobber.yaml
@@ -1,3 +1,4 @@
+# skip: redirect type is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Clobber redirection (>|) is not supported.
 input:
diff --git a/tests/scenarios/shell/blocked_redirects/write_truncate.yaml b/tests/scenarios/shell/blocked_redirects/write_truncate.yaml
index 0e18a4ef..0498ccb4 100644
--- a/tests/scenarios/shell/blocked_redirects/write_truncate.yaml
+++ b/tests/scenarios/shell/blocked_redirects/write_truncate.yaml
@@ -1,3 +1,4 @@
+# skip: redirect type is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Output redirection (>) is not supported.
 input:
diff --git a/tests/scenarios/shell/brace_group/basic/effect_of_brace.yaml b/tests/scenarios/shell/brace_group/basic/effect_of_brace.yaml
new file mode 100644
index 00000000..3e94a79e
--- /dev/null
+++ b/tests/scenarios/shell/brace_group/basic/effect_of_brace.yaml
@@ -0,0 +1,12 @@
+description: Brace group executes commands and variable assignments persist.
+input:
+  script: |+
+    a=1
+    { a=2; echo $a; }
+    echo $a
+expect:
+  stdout: |+
+    2
+    2
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/brace_group/basic/exit_status.yaml b/tests/scenarios/shell/brace_group/basic/exit_status.yaml
new file mode 100644
index 00000000..98f8dcfb
--- /dev/null
+++ b/tests/scenarios/shell/brace_group/basic/exit_status.yaml
@@ -0,0 +1,13 @@
+description: Brace group exit status reflects the last command.
+input:
+  script: |+
+    { true; false; }
+    echo $?
+    { false; true; }
+    echo $?
+expect:
+  stdout: |+
+    1
+    0
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/brace_group/basic/newlines.yaml b/tests/scenarios/shell/brace_group/basic/newlines.yaml
new file mode 100644
index 00000000..8ddd00a4
--- /dev/null
+++ b/tests/scenarios/shell/brace_group/basic/newlines.yaml
@@ -0,0 +1,11 @@
+description: Brace group with newline-separated commands.
+input:
+  script: |+
+    {
+    echo foo
+    }
+expect:
+  stdout: |+
+    foo
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/brace_group/basic/pipe_brace_output.yaml b/tests/scenarios/shell/brace_group/basic/pipe_brace_output.yaml
new file mode 100644
index 00000000..c8e2cbec
--- /dev/null
+++ b/tests/scenarios/shell/brace_group/basic/pipe_brace_output.yaml
@@ -0,0 +1,9 @@
+description: Brace group output can be piped.
+input:
+  script: |+
+    { echo 1; echo 2; echo 3; } | tail -n 1
+expect:
+  stdout: |+
+    3
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/brace_group/basic/semicolon_ending.yaml b/tests/scenarios/shell/brace_group/basic/semicolon_ending.yaml
new file mode 100644
index 00000000..8d3a083d
--- /dev/null
+++ b/tests/scenarios/shell/brace_group/basic/semicolon_ending.yaml
@@ -0,0 +1,9 @@
+description: Brace group ending with semicolon before closing brace.
+input:
+  script: |+
+    { echo foo; }
+expect:
+  stdout: |+
+    foo
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/brace_group/nested_subshell_blocked.yaml b/tests/scenarios/shell/brace_group/nested_subshell_blocked.yaml
index b3b9b3ae..fd27a576 100644
--- a/tests/scenarios/shell/brace_group/nested_subshell_blocked.yaml
+++ b/tests/scenarios/shell/brace_group/nested_subshell_blocked.yaml
@@ -1,3 +1,4 @@
+# skip: subshells are intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Subshell after subshell is blocked.
 input:
diff --git a/tests/scenarios/shell/brace_group/subshell_blocked.yaml b/tests/scenarios/shell/brace_group/subshell_blocked.yaml
index 6560dbb6..936da2b9 100644
--- a/tests/scenarios/shell/brace_group/subshell_blocked.yaml
+++ b/tests/scenarios/shell/brace_group/subshell_blocked.yaml
@@ -1,3 +1,4 @@
+# skip: subshells are intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Subshell grouping is blocked.
 input:
diff --git a/tests/scenarios/shell/case_clause/multiple_patterns.yaml b/tests/scenarios/shell/case_clause/multiple_patterns.yaml
index abdcb528..066535b6 100644
--- a/tests/scenarios/shell/case_clause/multiple_patterns.yaml
+++ b/tests/scenarios/shell/case_clause/multiple_patterns.yaml
@@ -1,3 +1,4 @@
+# skip: case statements are intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Case statement with multiple patterns is blocked.
 input:
diff --git a/tests/scenarios/shell/case_clause/paren_pattern.yaml b/tests/scenarios/shell/case_clause/paren_pattern.yaml
index 96568b70..fa29262d 100644
--- a/tests/scenarios/shell/case_clause/paren_pattern.yaml
+++ b/tests/scenarios/shell/case_clause/paren_pattern.yaml
@@ -1,3 +1,4 @@
+# skip: case statements are intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Case statement with parenthesized pattern is blocked.
 input:
diff --git a/tests/scenarios/shell/case_clause/wildcard_pattern.yaml b/tests/scenarios/shell/case_clause/wildcard_pattern.yaml
index aca88456..e81ae64c 100644
--- a/tests/scenarios/shell/case_clause/wildcard_pattern.yaml
+++ b/tests/scenarios/shell/case_clause/wildcard_pattern.yaml
@@ -1,3 +1,4 @@
+# skip: case statements are intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Case statement with wildcard pattern is blocked.
 input:
diff --git a/tests/scenarios/shell/environment/empty_by_default.yaml b/tests/scenarios/shell/environment/empty_by_default.yaml
index a6ac2378..c9872e19 100644
--- a/tests/scenarios/shell/environment/empty_by_default.yaml
+++ b/tests/scenarios/shell/environment/empty_by_default.yaml
@@ -1,3 +1,4 @@
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 description: Without input.envs, no environment variables are available except builtins.
 input:
diff --git a/tests/scenarios/shell/environment/env_option_empty_value.yaml b/tests/scenarios/shell/environment/env_option_empty_value.yaml
index b2d7a7e4..c496cdd0 100644
--- a/tests/scenarios/shell/environment/env_option_empty_value.yaml
+++ b/tests/scenarios/shell/environment/env_option_empty_value.yaml
@@ -1,4 +1,5 @@
 description: "Env option can provide empty-string values"
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 input:
   interpreter_env:
diff --git a/tests/scenarios/shell/environment/env_option_field_splitting.yaml b/tests/scenarios/shell/environment/env_option_field_splitting.yaml
index 4ba19d3e..0d8686e5 100644
--- a/tests/scenarios/shell/environment/env_option_field_splitting.yaml
+++ b/tests/scenarios/shell/environment/env_option_field_splitting.yaml
@@ -1,4 +1,5 @@
 description: "Env option provides variables used in field splitting"
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 input:
   interpreter_env:
diff --git a/tests/scenarios/shell/environment/env_option_no_extra_vars.yaml b/tests/scenarios/shell/environment/env_option_no_extra_vars.yaml
index 6635e483..8a4ba1f0 100644
--- a/tests/scenarios/shell/environment/env_option_no_extra_vars.yaml
+++ b/tests/scenarios/shell/environment/env_option_no_extra_vars.yaml
@@ -1,4 +1,5 @@
 description: "Variables from Env option do not pollute unrelated variable names"
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 input:
   interpreter_env:
diff --git a/tests/scenarios/shell/environment/env_option_override.yaml b/tests/scenarios/shell/environment/env_option_override.yaml
index 259885fa..da7aa6ab 100644
--- a/tests/scenarios/shell/environment/env_option_override.yaml
+++ b/tests/scenarios/shell/environment/env_option_override.yaml
@@ -1,4 +1,5 @@
 description: "Variables from Env option can be overridden by script assignment"
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 input:
   interpreter_env:
diff --git a/tests/scenarios/shell/environment/env_option_path_like_value.yaml b/tests/scenarios/shell/environment/env_option_path_like_value.yaml
index 15beda9f..85c1b9a5 100644
--- a/tests/scenarios/shell/environment/env_option_path_like_value.yaml
+++ b/tests/scenarios/shell/environment/env_option_path_like_value.yaml
@@ -1,4 +1,5 @@
 description: "Env option with PATH-like value containing colons"
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 input:
   interpreter_env:
diff --git a/tests/scenarios/shell/environment/env_option_special_chars.yaml b/tests/scenarios/shell/environment/env_option_special_chars.yaml
index 74b14018..0a1e1150 100644
--- a/tests/scenarios/shell/environment/env_option_special_chars.yaml
+++ b/tests/scenarios/shell/environment/env_option_special_chars.yaml
@@ -1,4 +1,5 @@
 description: "Env option variable with special characters in value"
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 input:
   interpreter_env:
diff --git a/tests/scenarios/shell/environment/env_option_vars_accessible.yaml b/tests/scenarios/shell/environment/env_option_vars_accessible.yaml
index f5405d79..7fbf646b 100644
--- a/tests/scenarios/shell/environment/env_option_vars_accessible.yaml
+++ b/tests/scenarios/shell/environment/env_option_vars_accessible.yaml
@@ -1,4 +1,5 @@
 description: "Variables provided via interpreter_env (Env option) are accessible"
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 input:
   interpreter_env:
diff --git a/tests/scenarios/shell/environment/home_not_set.yaml b/tests/scenarios/shell/environment/home_not_set.yaml
index 4895e9bf..93531a93 100644
--- a/tests/scenarios/shell/environment/home_not_set.yaml
+++ b/tests/scenarios/shell/environment/home_not_set.yaml
@@ -1,3 +1,4 @@
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 description: $HOME is not automatically set by the interpreter.
 input:
diff --git a/tests/scenarios/shell/environment/lang_not_set.yaml b/tests/scenarios/shell/environment/lang_not_set.yaml
index a556e3a5..af7ccb41 100644
--- a/tests/scenarios/shell/environment/lang_not_set.yaml
+++ b/tests/scenarios/shell/environment/lang_not_set.yaml
@@ -1,4 +1,5 @@
 description: "LANG variable is not set in the default environment"
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 input:
   script: |
diff --git a/tests/scenarios/shell/environment/no_parent_propagation.yaml b/tests/scenarios/shell/environment/no_parent_propagation.yaml
index 69a6721e..a4e1ac0d 100644
--- a/tests/scenarios/shell/environment/no_parent_propagation.yaml
+++ b/tests/scenarios/shell/environment/no_parent_propagation.yaml
@@ -1,3 +1,4 @@
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 description: Parent environment variables set via input.envs are not visible in the interpreter.
 input:
diff --git a/tests/scenarios/shell/environment/override_provided.yaml b/tests/scenarios/shell/environment/override_provided.yaml
index e358e464..979e8466 100644
--- a/tests/scenarios/shell/environment/override_provided.yaml
+++ b/tests/scenarios/shell/environment/override_provided.yaml
@@ -1,3 +1,4 @@
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 description: Script can assign variables that exist as parent env vars without inheriting them.
 input:
diff --git a/tests/scenarios/shell/environment/path_not_set.yaml b/tests/scenarios/shell/environment/path_not_set.yaml
index c3b8513b..7141d806 100644
--- a/tests/scenarios/shell/environment/path_not_set.yaml
+++ b/tests/scenarios/shell/environment/path_not_set.yaml
@@ -1,4 +1,5 @@
 description: "PATH variable is not set in the default environment"
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 input:
   script: |
diff --git a/tests/scenarios/shell/environment/provided_vars_accessible.yaml b/tests/scenarios/shell/environment/provided_vars_accessible.yaml
index 17f85a1c..cbf9b9e7 100644
--- a/tests/scenarios/shell/environment/provided_vars_accessible.yaml
+++ b/tests/scenarios/shell/environment/provided_vars_accessible.yaml
@@ -1,3 +1,4 @@
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 description: Even common parent env vars like PATH are not propagated to the interpreter.
 input:
diff --git a/tests/scenarios/shell/environment/shell_not_set.yaml b/tests/scenarios/shell/environment/shell_not_set.yaml
index d25602bb..8851b8e6 100644
--- a/tests/scenarios/shell/environment/shell_not_set.yaml
+++ b/tests/scenarios/shell/environment/shell_not_set.yaml
@@ -1,4 +1,5 @@
 description: "SHELL variable is not set in the default environment"
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 input:
   script: |
diff --git a/tests/scenarios/shell/environment/term_not_set.yaml b/tests/scenarios/shell/environment/term_not_set.yaml
index 9513e2be..0ce5070d 100644
--- a/tests/scenarios/shell/environment/term_not_set.yaml
+++ b/tests/scenarios/shell/environment/term_not_set.yaml
@@ -1,4 +1,5 @@
 description: "TERM variable is not set in the default environment"
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 input:
   script: |
diff --git a/tests/scenarios/shell/environment/tilde_in_heredoc_allowed.yaml b/tests/scenarios/shell/environment/tilde_in_heredoc_allowed.yaml
index 086d6c50..ab5ac5dc 100644
--- a/tests/scenarios/shell/environment/tilde_in_heredoc_allowed.yaml
+++ b/tests/scenarios/shell/environment/tilde_in_heredoc_allowed.yaml
@@ -1,3 +1,4 @@
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 description: Tilde inside heredoc body is allowed (heredocs do not undergo tilde expansion).
 input:
diff --git a/tests/scenarios/shell/environment/tilde_in_variable_assignment_blocked.yaml b/tests/scenarios/shell/environment/tilde_in_variable_assignment_blocked.yaml
index d28eafac..40e3f9e1 100644
--- a/tests/scenarios/shell/environment/tilde_in_variable_assignment_blocked.yaml
+++ b/tests/scenarios/shell/environment/tilde_in_variable_assignment_blocked.yaml
@@ -1,3 +1,4 @@
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 description: Tilde in variable assignment is blocked.
 input:
diff --git a/tests/scenarios/shell/environment/tilde_mid_word_allowed.yaml b/tests/scenarios/shell/environment/tilde_mid_word_allowed.yaml
index 68035c8d..b3cfcea0 100644
--- a/tests/scenarios/shell/environment/tilde_mid_word_allowed.yaml
+++ b/tests/scenarios/shell/environment/tilde_mid_word_allowed.yaml
@@ -1,3 +1,4 @@
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 description: Tilde in the middle of a word is not tilde expansion and should be allowed.
 input:
diff --git a/tests/scenarios/shell/environment/tilde_not_expanded.yaml b/tests/scenarios/shell/environment/tilde_not_expanded.yaml
index 330789a5..64055e82 100644
--- a/tests/scenarios/shell/environment/tilde_not_expanded.yaml
+++ b/tests/scenarios/shell/environment/tilde_not_expanded.yaml
@@ -1,3 +1,4 @@
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 description: Tilde expansion is blocked at validation with exit code 2.
 input:
diff --git a/tests/scenarios/shell/environment/tilde_path_not_expanded.yaml b/tests/scenarios/shell/environment/tilde_path_not_expanded.yaml
index 7e1bb4a1..d795f371 100644
--- a/tests/scenarios/shell/environment/tilde_path_not_expanded.yaml
+++ b/tests/scenarios/shell/environment/tilde_path_not_expanded.yaml
@@ -1,3 +1,4 @@
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 description: Tilde with path ~/dir is blocked at validation.
 input:
diff --git a/tests/scenarios/shell/environment/tilde_quoted_allowed.yaml b/tests/scenarios/shell/environment/tilde_quoted_allowed.yaml
index 1c5e8261..050eefa6 100644
--- a/tests/scenarios/shell/environment/tilde_quoted_allowed.yaml
+++ b/tests/scenarios/shell/environment/tilde_quoted_allowed.yaml
@@ -1,3 +1,4 @@
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 description: Tilde inside quotes is not tilde expansion and should be allowed.
 input:
diff --git a/tests/scenarios/shell/environment/tilde_username_blocked.yaml b/tests/scenarios/shell/environment/tilde_username_blocked.yaml
index 9275e4bb..76db6edc 100644
--- a/tests/scenarios/shell/environment/tilde_username_blocked.yaml
+++ b/tests/scenarios/shell/environment/tilde_username_blocked.yaml
@@ -1,3 +1,4 @@
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 description: Tilde with username (~root) is blocked to prevent host user info disclosure.
 input:
diff --git a/tests/scenarios/shell/environment/user_not_set.yaml b/tests/scenarios/shell/environment/user_not_set.yaml
index d1dd0cab..29246e20 100644
--- a/tests/scenarios/shell/environment/user_not_set.yaml
+++ b/tests/scenarios/shell/environment/user_not_set.yaml
@@ -1,4 +1,5 @@
 description: "USER variable is not set in the default environment"
+# skip: rshell uses an isolated environment that does not inherit from the parent process
 skip_assert_against_bash: true
 input:
   script: |
diff --git a/tests/scenarios/shell/errors/command_not_found.yaml b/tests/scenarios/shell/errors/command_not_found.yaml
index e509f86e..e75f8b09 100644
--- a/tests/scenarios/shell/errors/command_not_found.yaml
+++ b/tests/scenarios/shell/errors/command_not_found.yaml
@@ -1,10 +1,10 @@
-skip_assert_against_bash: true
 description: Unknown command returns exit code 127.
 input:
   script: |+
     no_such_command_xyz
 expect:
   stdout: ""
-  stderr: |+
-    no_such_command_xyz: command not found
+  stderr_contains:
+    - "no_such_command_xyz"
+    - "command not found"
   exit_code: 127
diff --git a/tests/scenarios/shell/errors/command_not_found_in_pipeline.yaml b/tests/scenarios/shell/errors/command_not_found_in_pipeline.yaml
index f1e2c1e6..5fc952c8 100644
--- a/tests/scenarios/shell/errors/command_not_found_in_pipeline.yaml
+++ b/tests/scenarios/shell/errors/command_not_found_in_pipeline.yaml
@@ -1,5 +1,3 @@
-# skip: command-not-found error format differs from bash (no script:line prefix)
-skip_assert_against_bash: true
 description: Unknown command in a pipeline produces error but pipeline continues.
 input:
   script: |+
@@ -7,5 +5,6 @@ input:
 expect:
   stdout: ""
   stderr_contains:
-    - "unknown_filter_cmd: command not found"
+    - "unknown_filter_cmd"
+    - "command not found"
   exit_code: 127
diff --git a/tests/scenarios/shell/errors/syntax_error_kills_shell.yaml b/tests/scenarios/shell/errors/syntax_error_kills_shell.yaml
index 2a8532e8..09061437 100644
--- a/tests/scenarios/shell/errors/syntax_error_kills_shell.yaml
+++ b/tests/scenarios/shell/errors/syntax_error_kills_shell.yaml
@@ -1,3 +1,4 @@
+# skip: error message format differs from bash (rshell uses different command-not-found format)
 skip_assert_against_bash: true
 description: Unknown command after valid command still produces error.
 input:
diff --git a/tests/scenarios/shell/field_splitting/consecutive_nonwhitespace_ifs.yaml b/tests/scenarios/shell/field_splitting/consecutive_nonwhitespace_ifs.yaml
index 716c0df7..d2671a6d 100644
--- a/tests/scenarios/shell/field_splitting/consecutive_nonwhitespace_ifs.yaml
+++ b/tests/scenarios/shell/field_splitting/consecutive_nonwhitespace_ifs.yaml
@@ -3,6 +3,7 @@ description: >
   In bash, "a::b" with IFS=: produces three fields: "a", "", "b".
   The upstream mvdan.cc/sh library does not produce the empty middle field.
   skip_assert_against_bash because of this known upstream limitation.
+# skip: known upstream library limitation with consecutive non-whitespace IFS delimiters
 skip_assert_against_bash: true
 input:
   script: |+
diff --git a/tests/scenarios/shell/field_splitting/ifs_behavior/empty_field_removal.yaml b/tests/scenarios/shell/field_splitting/ifs_behavior/empty_field_removal.yaml
new file mode 100644
index 00000000..ab7eb684
--- /dev/null
+++ b/tests/scenarios/shell/field_splitting/ifs_behavior/empty_field_removal.yaml
@@ -0,0 +1,13 @@
+description: Empty unquoted expansion produces no fields.
+input:
+  script: |+
+    a=
+    echo "[$a]"
+    for word in $a; do echo "word: $word"; done
+    echo done
+expect:
+  stdout: |+
+    []
+    done
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/field_splitting/ifs_behavior/no_split_empty_ifs.yaml b/tests/scenarios/shell/field_splitting/ifs_behavior/no_split_empty_ifs.yaml
new file mode 100644
index 00000000..2c859821
--- /dev/null
+++ b/tests/scenarios/shell/field_splitting/ifs_behavior/no_split_empty_ifs.yaml
@@ -0,0 +1,11 @@
+description: Empty IFS prevents all field splitting.
+input:
+  script: |+
+    a="one two three"
+    IFS=
+    echo $a
+expect:
+  stdout: |+
+    one two three
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/field_splitting/ifs_behavior/nonwhitespace_ifs.yaml b/tests/scenarios/shell/field_splitting/ifs_behavior/nonwhitespace_ifs.yaml
new file mode 100644
index 00000000..0c2759b6
--- /dev/null
+++ b/tests/scenarios/shell/field_splitting/ifs_behavior/nonwhitespace_ifs.yaml
@@ -0,0 +1,13 @@
+description: Non-whitespace IFS characters split fields.
+input:
+  script: |+
+    a="one-two-three"
+    IFS=-
+    for word in $a; do echo "[$word]"; done
+expect:
+  stdout: |+
+    [one]
+    [two]
+    [three]
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/field_splitting/ifs_behavior/quoted_no_split.yaml b/tests/scenarios/shell/field_splitting/ifs_behavior/quoted_no_split.yaml
new file mode 100644
index 00000000..b64a5193
--- /dev/null
+++ b/tests/scenarios/shell/field_splitting/ifs_behavior/quoted_no_split.yaml
@@ -0,0 +1,10 @@
+description: Quoted variable expansion prevents field splitting.
+input:
+  script: |+
+    a="one two three"
+    echo "$a"
+expect:
+  stdout: |+
+    one two three
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/field_splitting/ifs_behavior/standard_ifs.yaml b/tests/scenarios/shell/field_splitting/ifs_behavior/standard_ifs.yaml
new file mode 100644
index 00000000..83478078
--- /dev/null
+++ b/tests/scenarios/shell/field_splitting/ifs_behavior/standard_ifs.yaml
@@ -0,0 +1,12 @@
+description: Default IFS splits on space, tab, and newline.
+input:
+  script: |+
+    a="one two three"
+    for word in $a; do echo "[$word]"; done
+expect:
+  stdout: |+
+    [one]
+    [two]
+    [three]
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/break_cont/advanced/break_after_and.yaml b/tests/scenarios/shell/for_clause/break_cont/advanced/break_after_and.yaml
new file mode 100644
index 00000000..2e439ab5
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/break_cont/advanced/break_after_and.yaml
@@ -0,0 +1,13 @@
+description: Break after && operator exits the for loop.
+input:
+  script: |+
+    for i in 1; do
+        true && break
+        echo not reached
+    done
+    echo done
+expect:
+  stdout: |+
+    done
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/break_cont/advanced/break_after_or.yaml b/tests/scenarios/shell/for_clause/break_cont/advanced/break_after_or.yaml
new file mode 100644
index 00000000..88b4f2e1
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/break_cont/advanced/break_after_or.yaml
@@ -0,0 +1,13 @@
+description: Break after || operator exits the for loop.
+input:
+  script: |+
+    for i in 1; do
+        false || break
+        echo not reached
+    done
+    echo done
+expect:
+  stdout: |+
+    done
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/break_cont/advanced/break_default_operand.yaml b/tests/scenarios/shell/for_clause/break_cont/advanced/break_default_operand.yaml
new file mode 100644
index 00000000..ff6c93ad
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/break_cont/advanced/break_default_operand.yaml
@@ -0,0 +1,27 @@
+description: Break without operand defaults to break 1.
+input:
+  script: |+
+    for i in 1; do
+        echo in $i
+        for j in a; do
+            echo in $i $j
+            for k in x; do
+                echo in $i $j $k
+                break
+                echo out $i $j $k
+            done
+            echo out $i $j
+        done
+        echo out $i
+    done
+    echo done $?
+expect:
+  stdout: |+
+    in 1
+    in 1 a
+    in 1 a x
+    out 1 a
+    out 1
+    done 0
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/break_cont/advanced/break_nested_inner.yaml b/tests/scenarios/shell/for_clause/break_cont/advanced/break_nested_inner.yaml
new file mode 100644
index 00000000..d202c0d8
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/break_cont/advanced/break_nested_inner.yaml
@@ -0,0 +1,27 @@
+description: Break exits only the inner for loop.
+input:
+  script: |+
+    for i in 1 2 3; do
+        echo in $i
+        for j in a b c; do
+            echo in $i $j
+            break 1
+            echo out $i $j
+        done
+        echo out $i
+    done
+    echo done $?
+expect:
+  stdout: |+
+    in 1
+    in 1 a
+    out 1
+    in 2
+    in 2 a
+    out 2
+    in 3
+    in 3 a
+    out 3
+    done 0
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/break_cont/advanced/break_one_for.yaml b/tests/scenarios/shell/for_clause/break_cont/advanced/break_one_for.yaml
new file mode 100644
index 00000000..28f12c3a
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/break_cont/advanced/break_one_for.yaml
@@ -0,0 +1,15 @@
+description: Break exits the innermost for loop.
+input:
+  script: |+
+    for i in 1 2 3; do
+        echo in $i
+        break 1
+        echo out $i
+    done
+    echo done $?
+expect:
+  stdout: |+
+    in 1
+    done 0
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/break_cont/advanced/break_out_of_brace.yaml b/tests/scenarios/shell/for_clause/break_cont/advanced/break_out_of_brace.yaml
new file mode 100644
index 00000000..cc4584a4
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/break_cont/advanced/break_out_of_brace.yaml
@@ -0,0 +1,13 @@
+description: Break inside a brace group still exits the enclosing for loop.
+input:
+  script: |+
+    for i in 1; do
+        { break; }
+        echo not reached
+    done
+    echo done
+expect:
+  stdout: |+
+    done
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/break_cont/advanced/break_out_of_if.yaml b/tests/scenarios/shell/for_clause/break_cont/advanced/break_out_of_if.yaml
new file mode 100644
index 00000000..3dd8931e
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/break_cont/advanced/break_out_of_if.yaml
@@ -0,0 +1,13 @@
+description: Break inside an if-then exits the enclosing for loop.
+input:
+  script: |+
+    for i in 1; do
+        if true; then break; echo not reached then; fi
+        echo not reached
+    done
+    echo done
+expect:
+  stdout: |+
+    done
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/break_cont/advanced/break_two_levels.yaml b/tests/scenarios/shell/for_clause/break_cont/advanced/break_two_levels.yaml
new file mode 100644
index 00000000..a07a3663
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/break_cont/advanced/break_two_levels.yaml
@@ -0,0 +1,20 @@
+description: Break 2 exits both nested for loops.
+input:
+  script: |+
+    for i in 1 2 3; do
+        echo in $i
+        for j in a b c; do
+            echo in $i $j
+            break 2
+            echo out $i $j
+        done
+        echo out $i
+    done
+    echo done $?
+expect:
+  stdout: |+
+    in 1
+    in 1 a
+    done 0
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/break_cont/advanced/continue_after_and.yaml b/tests/scenarios/shell/for_clause/break_cont/advanced/continue_after_and.yaml
new file mode 100644
index 00000000..a665f4fd
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/break_cont/advanced/continue_after_and.yaml
@@ -0,0 +1,13 @@
+description: Continue after && operator continues the for loop.
+input:
+  script: |+
+    for i in 1 2; do
+        true && continue
+        echo not reached
+    done
+    echo done
+expect:
+  stdout: |+
+    done
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/break_cont/advanced/continue_after_or.yaml b/tests/scenarios/shell/for_clause/break_cont/advanced/continue_after_or.yaml
new file mode 100644
index 00000000..3b9c5aab
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/break_cont/advanced/continue_after_or.yaml
@@ -0,0 +1,13 @@
+description: Continue after || operator continues the for loop.
+input:
+  script: |+
+    for i in 1 2; do
+        false || continue
+        echo not reached
+    done
+    echo done
+expect:
+  stdout: |+
+    done
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/break_cont/advanced/continue_default_operand.yaml b/tests/scenarios/shell/for_clause/break_cont/advanced/continue_default_operand.yaml
new file mode 100644
index 00000000..5576a9c0
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/break_cont/advanced/continue_default_operand.yaml
@@ -0,0 +1,28 @@
+description: Continue without operand defaults to continue 1.
+input:
+  script: |+
+    for i in 1; do
+        echo in $i
+        for j in a; do
+            echo in $i $j
+            for k in x y; do
+                echo in $i $j $k
+                continue
+                echo out $i $j $k
+            done
+            echo out $i $j
+        done
+        echo out $i
+    done
+    echo done $?
+expect:
+  stdout: |+
+    in 1
+    in 1 a
+    in 1 a x
+    in 1 a y
+    out 1 a
+    out 1
+    done 0
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/break_cont/advanced/continue_nested_inner.yaml b/tests/scenarios/shell/for_clause/break_cont/advanced/continue_nested_inner.yaml
new file mode 100644
index 00000000..d6cc0d9a
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/break_cont/advanced/continue_nested_inner.yaml
@@ -0,0 +1,33 @@
+description: Continue in inner loop skips only inner loop body.
+input:
+  script: |+
+    for i in 1 2 3; do
+        echo in $i
+        for j in a b c; do
+            echo in $i $j
+            continue 1
+            echo out $i $j
+        done
+        echo out $i
+    done
+    echo done $?
+expect:
+  stdout: |+
+    in 1
+    in 1 a
+    in 1 b
+    in 1 c
+    out 1
+    in 2
+    in 2 a
+    in 2 b
+    in 2 c
+    out 2
+    in 3
+    in 3 a
+    in 3 b
+    in 3 c
+    out 3
+    done 0
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/break_cont/advanced/continue_one_for.yaml b/tests/scenarios/shell/for_clause/break_cont/advanced/continue_one_for.yaml
new file mode 100644
index 00000000..c9a04926
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/break_cont/advanced/continue_one_for.yaml
@@ -0,0 +1,17 @@
+description: Continue skips the rest of the loop body.
+input:
+  script: |+
+    for i in 1 2 3; do
+        echo in $i
+        continue 1
+        echo out $i
+    done
+    echo done $?
+expect:
+  stdout: |+
+    in 1
+    in 2
+    in 3
+    done 0
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/break_cont/advanced/continue_out_of_brace.yaml b/tests/scenarios/shell/for_clause/break_cont/advanced/continue_out_of_brace.yaml
new file mode 100644
index 00000000..f9fd8e05
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/break_cont/advanced/continue_out_of_brace.yaml
@@ -0,0 +1,13 @@
+description: Continue inside a brace group still affects the enclosing for loop.
+input:
+  script: |+
+    for i in 1 2; do
+        { continue; }
+        echo not reached
+    done
+    echo done
+expect:
+  stdout: |+
+    done
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/break_cont/advanced/continue_out_of_if.yaml b/tests/scenarios/shell/for_clause/break_cont/advanced/continue_out_of_if.yaml
new file mode 100644
index 00000000..7168bbd1
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/break_cont/advanced/continue_out_of_if.yaml
@@ -0,0 +1,13 @@
+description: Continue inside an if-then continues the enclosing for loop.
+input:
+  script: |+
+    for i in 1 2; do
+        if true; then continue; echo not reached then; fi
+        echo not reached
+    done
+    echo done
+expect:
+  stdout: |+
+    done
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/break_cont/advanced/continue_two_levels.yaml b/tests/scenarios/shell/for_clause/break_cont/advanced/continue_two_levels.yaml
new file mode 100644
index 00000000..252eab76
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/break_cont/advanced/continue_two_levels.yaml
@@ -0,0 +1,24 @@
+description: Continue 2 skips to the next iteration of the outer loop.
+input:
+  script: |+
+    for i in 1 2 3; do
+        echo in $i
+        for j in a b c; do
+            echo in $i $j
+            continue 2
+            echo out $i $j
+        done
+        echo out $i
+    done
+    echo done $?
+expect:
+  stdout: |+
+    in 1
+    in 1 a
+    in 2
+    in 2 a
+    in 3
+    in 3 a
+    done 0
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/break_cont/break_outside_loop.yaml b/tests/scenarios/shell/for_clause/break_cont/break_outside_loop.yaml
index bb52f1f0..bb22308b 100644
--- a/tests/scenarios/shell/for_clause/break_cont/break_outside_loop.yaml
+++ b/tests/scenarios/shell/for_clause/break_cont/break_outside_loop.yaml
@@ -1,3 +1,4 @@
+# skip: error message format differs from bash
 skip_assert_against_bash: true
 description: Break outside a loop produces an error message.
 input:
diff --git a/tests/scenarios/shell/for_clause/break_cont/continue_outside_loop.yaml b/tests/scenarios/shell/for_clause/break_cont/continue_outside_loop.yaml
index 7408ef41..4008f35b 100644
--- a/tests/scenarios/shell/for_clause/break_cont/continue_outside_loop.yaml
+++ b/tests/scenarios/shell/for_clause/break_cont/continue_outside_loop.yaml
@@ -1,3 +1,4 @@
+# skip: error message format differs from bash
 skip_assert_against_bash: true
 description: Continue outside a loop produces an error message.
 input:
diff --git a/tests/scenarios/shell/for_clause/edge_cases/exit_status_last_cmd.yaml b/tests/scenarios/shell/for_clause/edge_cases/exit_status_last_cmd.yaml
new file mode 100644
index 00000000..3df60d98
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/edge_cases/exit_status_last_cmd.yaml
@@ -0,0 +1,13 @@
+description: For loop exit status reflects last command in body.
+input:
+  script: |+
+    for i in 1 2 3; do true; done
+    echo $?
+    for i in 1 2 3; do false; done
+    echo $?
+expect:
+  stdout: |+
+    0
+    1
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/edge_cases/exit_status_no_words.yaml b/tests/scenarios/shell/for_clause/edge_cases/exit_status_no_words.yaml
new file mode 100644
index 00000000..ce3ade80
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/edge_cases/exit_status_no_words.yaml
@@ -0,0 +1,10 @@
+description: For loop with empty word list has exit status 0.
+input:
+  script: |+
+    for i in; do echo $i; done
+    echo $?
+expect:
+  stdout: |+
+    0
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/edge_cases/for_as_varname.yaml b/tests/scenarios/shell/for_clause/edge_cases/for_as_varname.yaml
new file mode 100644
index 00000000..8d507dd0
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/edge_cases/for_as_varname.yaml
@@ -0,0 +1,10 @@
+description: The word "for" can be used as the iteration variable name.
+input:
+  script: |+
+    for for in 1 2; do echo $for; done
+expect:
+  stdout: |+
+    1
+    2
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/edge_cases/iteration_var_global.yaml b/tests/scenarios/shell/for_clause/edge_cases/iteration_var_global.yaml
new file mode 100644
index 00000000..685a0923
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/edge_cases/iteration_var_global.yaml
@@ -0,0 +1,10 @@
+description: For loop iteration variable persists after loop completion.
+input:
+  script: |+
+    for myvar in a b c; do true; done
+    echo $myvar
+expect:
+  stdout: |+
+    c
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/edge_cases/pipe_for_output.yaml b/tests/scenarios/shell/for_clause/edge_cases/pipe_for_output.yaml
new file mode 100644
index 00000000..86060287
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/edge_cases/pipe_for_output.yaml
@@ -0,0 +1,11 @@
+description: For loop output can be piped.
+input:
+  script: |+
+    for i in 1 2 3; do echo $i; done | cat
+expect:
+  stdout: |+
+    1
+    2
+    3
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/edge_cases/semicolon_before_do.yaml b/tests/scenarios/shell/for_clause/edge_cases/semicolon_before_do.yaml
new file mode 100644
index 00000000..b9a6084b
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/edge_cases/semicolon_before_do.yaml
@@ -0,0 +1,11 @@
+description: Semicolon can separate the word list from do.
+input:
+  script: |+
+    for i in 1 2 3; do echo $i; done
+expect:
+  stdout: |+
+    1
+    2
+    3
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/for_clause/edge_cases/words_not_assignments.yaml b/tests/scenarios/shell/for_clause/edge_cases/words_not_assignments.yaml
new file mode 100644
index 00000000..1897ba4e
--- /dev/null
+++ b/tests/scenarios/shell/for_clause/edge_cases/words_not_assignments.yaml
@@ -0,0 +1,12 @@
+description: For loop words are not evaluated as variable assignments.
+input:
+  script: |+
+    for i in a=1 b=2; do echo $i; done
+    echo $a $b
+expect:
+  stdout: |+
+    a=1
+    b=2
+
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/function/function_for_body.yaml b/tests/scenarios/shell/function/function_for_body.yaml
index 6cdaef27..abb80407 100644
--- a/tests/scenarios/shell/function/function_for_body.yaml
+++ b/tests/scenarios/shell/function/function_for_body.yaml
@@ -1,3 +1,4 @@
+# skip: function declarations are intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Function with for loop body is blocked.
 input:
diff --git a/tests/scenarios/shell/function/function_portable_name.yaml b/tests/scenarios/shell/function/function_portable_name.yaml
index 8e65adfb..4f7bab40 100644
--- a/tests/scenarios/shell/function/function_portable_name.yaml
+++ b/tests/scenarios/shell/function/function_portable_name.yaml
@@ -1,3 +1,4 @@
+# skip: function declarations are intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Function with underscore and alphanumeric name is blocked.
 input:
diff --git a/tests/scenarios/shell/function/function_redefine.yaml b/tests/scenarios/shell/function/function_redefine.yaml
index 5736f664..dc67d72e 100644
--- a/tests/scenarios/shell/function/function_redefine.yaml
+++ b/tests/scenarios/shell/function/function_redefine.yaml
@@ -1,3 +1,4 @@
+# skip: function declarations are intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Re-defining a function is blocked.
 input:
diff --git a/tests/scenarios/shell/heredoc/basic/basic_heredoc.yaml b/tests/scenarios/shell/heredoc/basic/basic_heredoc.yaml
new file mode 100644
index 00000000..71536979
--- /dev/null
+++ b/tests/scenarios/shell/heredoc/basic/basic_heredoc.yaml
@@ -0,0 +1,15 @@
+description: Basic heredoc with content.
+input:
+  script: |+
+    cat <<END
+    here
+
+    	document
+    END
+expect:
+  stdout: |+
+    here
+
+    	document
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/heredoc/basic/delimiter_starting_with_dash.yaml b/tests/scenarios/shell/heredoc/basic/delimiter_starting_with_dash.yaml
new file mode 100644
index 00000000..0d487007
--- /dev/null
+++ b/tests/scenarios/shell/heredoc/basic/delimiter_starting_with_dash.yaml
@@ -0,0 +1,13 @@
+description: Heredoc delimiter can start with a dash character.
+input:
+  script: |+
+    cat << -END
+    foo
+    END
+    -END
+expect:
+  stdout: |+
+    foo
+    END
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/heredoc/basic/multiple_sequential.yaml b/tests/scenarios/shell/heredoc/basic/multiple_sequential.yaml
new file mode 100644
index 00000000..880cfa1f
--- /dev/null
+++ b/tests/scenarios/shell/heredoc/basic/multiple_sequential.yaml
@@ -0,0 +1,15 @@
+description: Multiple sequential commands each with their own heredoc.
+input:
+  script: |+
+    cat <<END1; echo ---; cat <<END2
+    first
+    END1
+    second
+    END2
+expect:
+  stdout: |+
+    first
+    ---
+    second
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/heredoc/basic/no_tilde_expansion.yaml b/tests/scenarios/shell/heredoc/basic/no_tilde_expansion.yaml
new file mode 100644
index 00000000..b1112ca7
--- /dev/null
+++ b/tests/scenarios/shell/heredoc/basic/no_tilde_expansion.yaml
@@ -0,0 +1,12 @@
+description: Tilde is not expanded in heredoc content.
+input:
+  script: |+
+    HOME=/home
+    cat <<END
+    tilde ~
+    END
+expect:
+  stdout: |+
+    tilde ~
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/heredoc/basic/quoted_delimiter_no_expansion.yaml b/tests/scenarios/shell/heredoc/basic/quoted_delimiter_no_expansion.yaml
new file mode 100644
index 00000000..76ce228b
--- /dev/null
+++ b/tests/scenarios/shell/heredoc/basic/quoted_delimiter_no_expansion.yaml
@@ -0,0 +1,12 @@
+description: Quoted heredoc delimiter prevents all expansions.
+input:
+  script: |+
+    foo=bar
+    cat <<'END'
+    parameter ${foo}
+    END
+expect:
+  stdout: |+
+    parameter ${foo}
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/heredoc/basic/single_double_quotes.yaml b/tests/scenarios/shell/heredoc/basic/single_double_quotes.yaml
new file mode 100644
index 00000000..8c289238
--- /dev/null
+++ b/tests/scenarios/shell/heredoc/basic/single_double_quotes.yaml
@@ -0,0 +1,11 @@
+description: Quotes in heredoc content are preserved literally.
+input:
+  script: |+
+    cat <<END
+    quote 'single' "double"
+    END
+expect:
+  stdout: |+
+    quote 'single' "double"
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/heredoc/basic/tab_removal.yaml b/tests/scenarios/shell/heredoc/basic/tab_removal.yaml
new file mode 100644
index 00000000..d10a117a
--- /dev/null
+++ b/tests/scenarios/shell/heredoc/basic/tab_removal.yaml
@@ -0,0 +1,9 @@
+description: Leading tabs are stripped with <<- operator.
+input:
+  script: "cat <<-END\n\tfoo\n\t\t\tbar\n\tEND\n"
+expect:
+  stdout: |+
+    foo
+    bar
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/heredoc/basic/var_expansion_unquoted.yaml b/tests/scenarios/shell/heredoc/basic/var_expansion_unquoted.yaml
new file mode 100644
index 00000000..2a759533
--- /dev/null
+++ b/tests/scenarios/shell/heredoc/basic/var_expansion_unquoted.yaml
@@ -0,0 +1,12 @@
+description: Variable expansion works in unquoted heredoc.
+input:
+  script: |+
+    foo=bar
+    cat <<END
+    parameter $foo
+    END
+expect:
+  stdout: |+
+    parameter bar
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/heredoc/basic/various_quoted_delimiter.yaml b/tests/scenarios/shell/heredoc/basic/various_quoted_delimiter.yaml
new file mode 100644
index 00000000..b52fac73
--- /dev/null
+++ b/tests/scenarios/shell/heredoc/basic/various_quoted_delimiter.yaml
@@ -0,0 +1,11 @@
+description: Heredoc delimiter can be partially quoted.
+input:
+  script: |+
+    cat <<E'N'D
+    single
+    END
+expect:
+  stdout: |+
+    single
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/if_clause/basic/elif_else_false_false.yaml b/tests/scenarios/shell/if_clause/basic/elif_else_false_false.yaml
new file mode 100644
index 00000000..37288816
--- /dev/null
+++ b/tests/scenarios/shell/if_clause/basic/elif_else_false_false.yaml
@@ -0,0 +1,9 @@
+description: Else branch executes when both if and elif fail.
+input:
+  script: |+
+    if false; then echo 1; elif false; then echo 2; else echo 3; fi
+expect:
+  stdout: |+
+    3
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/if_clause/basic/elif_else_false_true.yaml b/tests/scenarios/shell/if_clause/basic/elif_else_false_true.yaml
new file mode 100644
index 00000000..6731220a
--- /dev/null
+++ b/tests/scenarios/shell/if_clause/basic/elif_else_false_true.yaml
@@ -0,0 +1,9 @@
+description: Elif branch executes in if-elif-else when if fails and elif succeeds.
+input:
+  script: |+
+    if false; then echo 1; elif true; then echo 2; else echo 3; fi
+expect:
+  stdout: |+
+    2
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/if_clause/basic/elif_false_false.yaml b/tests/scenarios/shell/if_clause/basic/elif_false_false.yaml
new file mode 100644
index 00000000..20cb2887
--- /dev/null
+++ b/tests/scenarios/shell/if_clause/basic/elif_false_false.yaml
@@ -0,0 +1,10 @@
+description: Neither branch executes when both conditions fail.
+input:
+  script: |+
+    if false; then echo 1; elif false; then echo 2; fi
+    echo $?
+expect:
+  stdout: |+
+    0
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/if_clause/basic/elif_false_true.yaml b/tests/scenarios/shell/if_clause/basic/elif_false_true.yaml
new file mode 100644
index 00000000..4058a0aa
--- /dev/null
+++ b/tests/scenarios/shell/if_clause/basic/elif_false_true.yaml
@@ -0,0 +1,9 @@
+description: Elif branch executes when if condition fails.
+input:
+  script: |+
+    if false; then echo 1; elif true; then echo 2; fi
+expect:
+  stdout: |+
+    2
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/if_clause/basic/exit_status_true_then_false.yaml b/tests/scenarios/shell/if_clause/basic/exit_status_true_then_false.yaml
new file mode 100644
index 00000000..e1810608
--- /dev/null
+++ b/tests/scenarios/shell/if_clause/basic/exit_status_true_then_false.yaml
@@ -0,0 +1,10 @@
+description: Exit status of if reflects the body command when condition is true.
+input:
+  script: |+
+    if true; then false; fi
+    echo $?
+expect:
+  stdout: |+
+    1
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/if_clause/basic/exit_status_true_then_true.yaml b/tests/scenarios/shell/if_clause/basic/exit_status_true_then_true.yaml
new file mode 100644
index 00000000..dc7b7874
--- /dev/null
+++ b/tests/scenarios/shell/if_clause/basic/exit_status_true_then_true.yaml
@@ -0,0 +1,10 @@
+description: Exit status of if reflects the body command.
+input:
+  script: |+
+    if true; then true; fi
+    echo $?
+expect:
+  stdout: |+
+    0
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/if_clause/basic/if_else_false.yaml b/tests/scenarios/shell/if_clause/basic/if_else_false.yaml
new file mode 100644
index 00000000..87fc837d
--- /dev/null
+++ b/tests/scenarios/shell/if_clause/basic/if_else_false.yaml
@@ -0,0 +1,9 @@
+description: Else branch executes when if condition is false.
+input:
+  script: |+
+    if false; then echo then; else echo else; fi
+expect:
+  stdout: |+
+    else
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/if_clause/basic/if_else_true.yaml b/tests/scenarios/shell/if_clause/basic/if_else_true.yaml
new file mode 100644
index 00000000..70688001
--- /dev/null
+++ b/tests/scenarios/shell/if_clause/basic/if_else_true.yaml
@@ -0,0 +1,9 @@
+description: Then branch executes when if condition is true.
+input:
+  script: |+
+    if true; then echo then; else echo else; fi
+expect:
+  stdout: |+
+    then
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/if_clause/basic/if_false.yaml b/tests/scenarios/shell/if_clause/basic/if_false.yaml
new file mode 100644
index 00000000..ee286d09
--- /dev/null
+++ b/tests/scenarios/shell/if_clause/basic/if_false.yaml
@@ -0,0 +1,10 @@
+description: If block is skipped when condition is false.
+input:
+  script: |+
+    if false; then echo not reached; fi
+    echo $?
+expect:
+  stdout: |+
+    0
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/if_clause/basic/if_true.yaml b/tests/scenarios/shell/if_clause/basic/if_true.yaml
new file mode 100644
index 00000000..caf19abe
--- /dev/null
+++ b/tests/scenarios/shell/if_clause/basic/if_true.yaml
@@ -0,0 +1,9 @@
+description: If block executes when condition is true.
+input:
+  script: |+
+    if true; then echo reached; fi
+expect:
+  stdout: |+
+    reached
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/if_clause/basic/multiline_linebreaks.yaml b/tests/scenarios/shell/if_clause/basic/multiline_linebreaks.yaml
new file mode 100644
index 00000000..e415cab1
--- /dev/null
+++ b/tests/scenarios/shell/if_clause/basic/multiline_linebreaks.yaml
@@ -0,0 +1,13 @@
+description: Linebreaks are allowed in various positions within if statement.
+input:
+  script: |+
+    if
+    true
+    then
+    echo ok
+    fi
+expect:
+  stdout: |+
+    ok
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/if_clause/basic/nested_in_else.yaml b/tests/scenarios/shell/if_clause/basic/nested_in_else.yaml
new file mode 100644
index 00000000..e705d823
--- /dev/null
+++ b/tests/scenarios/shell/if_clause/basic/nested_in_else.yaml
@@ -0,0 +1,11 @@
+description: If statement nested inside else branch.
+input:
+  script: |+
+    if false; then echo 1; else
+      if true; then echo 2; fi
+    fi
+expect:
+  stdout: |+
+    2
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/if_clause/basic/nested_in_then.yaml b/tests/scenarios/shell/if_clause/basic/nested_in_then.yaml
new file mode 100644
index 00000000..22227ce6
--- /dev/null
+++ b/tests/scenarios/shell/if_clause/basic/nested_in_then.yaml
@@ -0,0 +1,11 @@
+description: If statement nested inside then branch.
+input:
+  script: |+
+    if true; then
+      if true; then echo inner; fi
+    fi
+expect:
+  stdout: |+
+    inner
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/if_clause/basic/pipe_if_output.yaml b/tests/scenarios/shell/if_clause/basic/pipe_if_output.yaml
new file mode 100644
index 00000000..5dcbb3e1
--- /dev/null
+++ b/tests/scenarios/shell/if_clause/basic/pipe_if_output.yaml
@@ -0,0 +1,9 @@
+description: If statement output can be piped.
+input:
+  script: |+
+    if true; then echo hello; fi | cat
+expect:
+  stdout: |+
+    hello
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/if_clause/edge_cases/blocked_in_condition.yaml b/tests/scenarios/shell/if_clause/edge_cases/blocked_in_condition.yaml
index 89fca5f0..295c0592 100644
--- a/tests/scenarios/shell/if_clause/edge_cases/blocked_in_condition.yaml
+++ b/tests/scenarios/shell/if_clause/edge_cases/blocked_in_condition.yaml
@@ -1,3 +1,4 @@
+# skip: tests blocked feature interaction with if clause (rshell-specific behavior)
 skip_assert_against_bash: true
 description: Blocked features inside if condition are still rejected.
 input:
diff --git a/tests/scenarios/shell/line_continuation/basic/in_assignment.yaml b/tests/scenarios/shell/line_continuation/basic/in_assignment.yaml
new file mode 100644
index 00000000..348c3c70
--- /dev/null
+++ b/tests/scenarios/shell/line_continuation/basic/in_assignment.yaml
@@ -0,0 +1,11 @@
+description: Backslash-newline works in variable assignments.
+input:
+  script: |+
+    fo\
+    o=bar
+    echo $foo
+expect:
+  stdout: |+
+    bar
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/line_continuation/basic/in_for_keyword.yaml b/tests/scenarios/shell/line_continuation/basic/in_for_keyword.yaml
new file mode 100644
index 00000000..1c968f75
--- /dev/null
+++ b/tests/scenarios/shell/line_continuation/basic/in_for_keyword.yaml
@@ -0,0 +1,17 @@
+description: Line continuation works across for/in/do/done keywords.
+input:
+  script: |+
+    f\
+    o\
+    r i i\
+    n 1 2; d\
+    o echo $i; d\
+    o\
+    n\
+    e
+expect:
+  stdout: |+
+    1
+    2
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/line_continuation/basic/in_if_keyword.yaml b/tests/scenarios/shell/line_continuation/basic/in_if_keyword.yaml
new file mode 100644
index 00000000..f767b046
--- /dev/null
+++ b/tests/scenarios/shell/line_continuation/basic/in_if_keyword.yaml
@@ -0,0 +1,14 @@
+description: Line continuation works across if/then/fi keywords.
+input:
+  script: |+
+    i\
+    f true; t\
+    h\
+    e\
+    n echo ok; f\
+    i
+expect:
+  stdout: |+
+    ok
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/line_continuation/basic/in_normal_word.yaml b/tests/scenarios/shell/line_continuation/basic/in_normal_word.yaml
new file mode 100644
index 00000000..95a1e6d9
--- /dev/null
+++ b/tests/scenarios/shell/line_continuation/basic/in_normal_word.yaml
@@ -0,0 +1,10 @@
+description: Backslash-newline joins adjacent lines into a single word.
+input:
+  script: |+
+    echo 123\
+    456
+expect:
+  stdout: |+
+    123456
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/logic_ops/and_or_chains/exit_status_last_executed.yaml b/tests/scenarios/shell/logic_ops/and_or_chains/exit_status_last_executed.yaml
new file mode 100644
index 00000000..f1cae6c9
--- /dev/null
+++ b/tests/scenarios/shell/logic_ops/and_or_chains/exit_status_last_executed.yaml
@@ -0,0 +1,19 @@
+description: Exit status reflects the last-executed pipeline in and-or list.
+input:
+  script: |+
+    true && false
+    echo $?
+    false && true
+    echo $?
+    true || false
+    echo $?
+    false || true
+    echo $?
+expect:
+  stdout: |+
+    1
+    1
+    0
+    0
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/logic_ops/and_or_chains/failure_and_failure.yaml b/tests/scenarios/shell/logic_ops/and_or_chains/failure_and_failure.yaml
new file mode 100644
index 00000000..b8925649
--- /dev/null
+++ b/tests/scenarios/shell/logic_ops/and_or_chains/failure_and_failure.yaml
@@ -0,0 +1,8 @@
+description: First command fails so second is skipped in && list.
+input:
+  script: |+
+    false && false
+expect:
+  stdout: ""
+  stderr: ""
+  exit_code: 1
diff --git a/tests/scenarios/shell/logic_ops/and_or_chains/failure_and_success.yaml b/tests/scenarios/shell/logic_ops/and_or_chains/failure_and_success.yaml
new file mode 100644
index 00000000..e6d3b504
--- /dev/null
+++ b/tests/scenarios/shell/logic_ops/and_or_chains/failure_and_success.yaml
@@ -0,0 +1,8 @@
+description: First command fails so second is skipped in && list.
+input:
+  script: |+
+    false && echo 2
+expect:
+  stdout: ""
+  stderr: ""
+  exit_code: 1
diff --git a/tests/scenarios/shell/logic_ops/and_or_chains/failure_or_failure.yaml b/tests/scenarios/shell/logic_ops/and_or_chains/failure_or_failure.yaml
new file mode 100644
index 00000000..6063a3ca
--- /dev/null
+++ b/tests/scenarios/shell/logic_ops/and_or_chains/failure_or_failure.yaml
@@ -0,0 +1,8 @@
+description: Both commands fail in || list, exit status is from last.
+input:
+  script: |+
+    false || false
+expect:
+  stdout: ""
+  stderr: ""
+  exit_code: 1
diff --git a/tests/scenarios/shell/logic_ops/and_or_chains/failure_or_success.yaml b/tests/scenarios/shell/logic_ops/and_or_chains/failure_or_success.yaml
new file mode 100644
index 00000000..fc2068c6
--- /dev/null
+++ b/tests/scenarios/shell/logic_ops/and_or_chains/failure_or_success.yaml
@@ -0,0 +1,9 @@
+description: First command fails so second runs in || list.
+input:
+  script: |+
+    false || echo 2
+expect:
+  stdout: |+
+    2
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/logic_ops/and_or_chains/linebreak_after_and.yaml b/tests/scenarios/shell/logic_ops/and_or_chains/linebreak_after_and.yaml
new file mode 100644
index 00000000..56d8a330
--- /dev/null
+++ b/tests/scenarios/shell/logic_ops/and_or_chains/linebreak_after_and.yaml
@@ -0,0 +1,13 @@
+description: Linebreak after && operator continues the and-or list.
+input:
+  script: |+
+    echo 1 &&
+    echo 2 &&
+    echo 3
+expect:
+  stdout: |+
+    1
+    2
+    3
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/logic_ops/and_or_chains/linebreak_after_or.yaml b/tests/scenarios/shell/logic_ops/and_or_chains/linebreak_after_or.yaml
new file mode 100644
index 00000000..b5dc2611
--- /dev/null
+++ b/tests/scenarios/shell/logic_ops/and_or_chains/linebreak_after_or.yaml
@@ -0,0 +1,11 @@
+description: Linebreak after || operator continues the and-or list.
+input:
+  script: |+
+    false ||
+    false ||
+    echo foo
+expect:
+  stdout: |+
+    foo
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/logic_ops/and_or_chains/success_and_failure.yaml b/tests/scenarios/shell/logic_ops/and_or_chains/success_and_failure.yaml
new file mode 100644
index 00000000..a2a77ff8
--- /dev/null
+++ b/tests/scenarios/shell/logic_ops/and_or_chains/success_and_failure.yaml
@@ -0,0 +1,9 @@
+description: First succeeds, second fails in && list.
+input:
+  script: |+
+    echo 1 && false
+expect:
+  stdout: |+
+    1
+  stderr: ""
+  exit_code: 1
diff --git a/tests/scenarios/shell/logic_ops/and_or_chains/success_and_success.yaml b/tests/scenarios/shell/logic_ops/and_or_chains/success_and_success.yaml
new file mode 100644
index 00000000..d57d4a49
--- /dev/null
+++ b/tests/scenarios/shell/logic_ops/and_or_chains/success_and_success.yaml
@@ -0,0 +1,10 @@
+description: Both commands in && list succeed and produce output.
+input:
+  script: |+
+    echo 1 && echo 2
+expect:
+  stdout: |+
+    1
+    2
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/logic_ops/and_or_chains/success_or_failure.yaml b/tests/scenarios/shell/logic_ops/and_or_chains/success_or_failure.yaml
new file mode 100644
index 00000000..0a8524c0
--- /dev/null
+++ b/tests/scenarios/shell/logic_ops/and_or_chains/success_or_failure.yaml
@@ -0,0 +1,9 @@
+description: First succeeds so second is skipped in || list.
+input:
+  script: |+
+    echo 1 || false
+expect:
+  stdout: |+
+    1
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/logic_ops/and_or_chains/success_or_success.yaml b/tests/scenarios/shell/logic_ops/and_or_chains/success_or_success.yaml
new file mode 100644
index 00000000..e855aa62
--- /dev/null
+++ b/tests/scenarios/shell/logic_ops/and_or_chains/success_or_success.yaml
@@ -0,0 +1,9 @@
+description: First command succeeds so second is skipped in || list.
+input:
+  script: |+
+    echo 1 || echo 2
+expect:
+  stdout: |+
+    1
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/logic_ops/and_or_chains/three_cmd_and_or_list.yaml b/tests/scenarios/shell/logic_ops/and_or_chains/three_cmd_and_or_list.yaml
new file mode 100644
index 00000000..8b32a006
--- /dev/null
+++ b/tests/scenarios/shell/logic_ops/and_or_chains/three_cmd_and_or_list.yaml
@@ -0,0 +1,11 @@
+description: Three-command and-or list with mixed && and ||.
+input:
+  script: |+
+    false && echo foo || echo bar
+    true || echo foo && echo bar
+expect:
+  stdout: |+
+    bar
+    bar
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/pipe/advanced/compound_in_pipeline.yaml b/tests/scenarios/shell/pipe/advanced/compound_in_pipeline.yaml
new file mode 100644
index 00000000..3197a2ae
--- /dev/null
+++ b/tests/scenarios/shell/pipe/advanced/compound_in_pipeline.yaml
@@ -0,0 +1,9 @@
+description: Brace groups and if statements can appear in a pipeline.
+input:
+  script: |+
+    { echo foo; echo bar; } | tail -n 1
+expect:
+  stdout: |+
+    bar
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/pipe/advanced/exit_status_last_cmd.yaml b/tests/scenarios/shell/pipe/advanced/exit_status_last_cmd.yaml
new file mode 100644
index 00000000..343f5b87
--- /dev/null
+++ b/tests/scenarios/shell/pipe/advanced/exit_status_last_cmd.yaml
@@ -0,0 +1,13 @@
+description: Pipeline exit status comes from the last command.
+input:
+  script: |+
+    echo a | true
+    echo a $?
+    echo b | false
+    echo b $?
+expect:
+  stdout: |+
+    a 0
+    b 1
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/pipe/advanced/linebreak_after_pipe.yaml b/tests/scenarios/shell/pipe/advanced/linebreak_after_pipe.yaml
new file mode 100644
index 00000000..be3ac29d
--- /dev/null
+++ b/tests/scenarios/shell/pipe/advanced/linebreak_after_pipe.yaml
@@ -0,0 +1,11 @@
+description: Linebreak after pipe operator continues the pipeline.
+input:
+  script: |+
+    printf '%s\n' foo bar |
+    tail -n 1 |
+    cat
+expect:
+  stdout: |+
+    bar
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/pipe/advanced/redirect_overrides_pipeline.yaml b/tests/scenarios/shell/pipe/advanced/redirect_overrides_pipeline.yaml
new file mode 100644
index 00000000..5cb43714
--- /dev/null
+++ b/tests/scenarios/shell/pipe/advanced/redirect_overrides_pipeline.yaml
@@ -0,0 +1,8 @@
+description: Explicit redirection to /dev/null overrides pipeline stdout.
+input:
+  script: |+
+    echo foo >/dev/null | cat
+expect:
+  stdout: ""
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/pipe/advanced/three_cmd_pipeline.yaml b/tests/scenarios/shell/pipe/advanced/three_cmd_pipeline.yaml
new file mode 100644
index 00000000..b830787f
--- /dev/null
+++ b/tests/scenarios/shell/pipe/advanced/three_cmd_pipeline.yaml
@@ -0,0 +1,9 @@
+description: Three-command pipeline passes data through.
+input:
+  script: |+
+    printf '%s\n' foo bar | tail -n 1 | cat
+expect:
+  stdout: |+
+    bar
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/pipe/advanced/two_cmd_pipeline.yaml b/tests/scenarios/shell/pipe/advanced/two_cmd_pipeline.yaml
new file mode 100644
index 00000000..ec63d756
--- /dev/null
+++ b/tests/scenarios/shell/pipe/advanced/two_cmd_pipeline.yaml
@@ -0,0 +1,9 @@
+description: Basic two-command pipeline.
+input:
+  script: |+
+    echo foo | cat
+expect:
+  stdout: |+
+    foo
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/readonly/blocked.yaml b/tests/scenarios/shell/readonly/blocked.yaml
index 10c41573..150a0f7f 100644
--- a/tests/scenarios/shell/readonly/blocked.yaml
+++ b/tests/scenarios/shell/readonly/blocked.yaml
@@ -1,3 +1,4 @@
+# skip: readonly is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: The readonly keyword is not supported and is blocked at parse validation.
 input:
diff --git a/tests/scenarios/shell/redirections/devnull/devnull_path_traversal_blocked.yaml b/tests/scenarios/shell/redirections/devnull/devnull_path_traversal_blocked.yaml
index fc1432b2..23eb8ec0 100644
--- a/tests/scenarios/shell/redirections/devnull/devnull_path_traversal_blocked.yaml
+++ b/tests/scenarios/shell/redirections/devnull/devnull_path_traversal_blocked.yaml
@@ -1,3 +1,4 @@
+# skip: redirect restrictions are an rshell-specific security feature
 skip_assert_against_bash: true
 description: Path traversal via /dev/null/../../etc is blocked (not literal /dev/null).
 input:
diff --git a/tests/scenarios/shell/redirections/devnull/redirect_to_file_still_blocked.yaml b/tests/scenarios/shell/redirections/devnull/redirect_to_file_still_blocked.yaml
index 2fcaa95b..d6bc61ae 100644
--- a/tests/scenarios/shell/redirections/devnull/redirect_to_file_still_blocked.yaml
+++ b/tests/scenarios/shell/redirections/devnull/redirect_to_file_still_blocked.yaml
@@ -1,3 +1,4 @@
+# skip: redirect restrictions are an rshell-specific security feature
 skip_assert_against_bash: true
 description: Output redirection to a real file (not /dev/null) is still blocked.
 input:
diff --git a/tests/scenarios/shell/redirections/devnull/stderr_redirect_to_file_blocked.yaml b/tests/scenarios/shell/redirections/devnull/stderr_redirect_to_file_blocked.yaml
index f0bebcd1..4ff81815 100644
--- a/tests/scenarios/shell/redirections/devnull/stderr_redirect_to_file_blocked.yaml
+++ b/tests/scenarios/shell/redirections/devnull/stderr_redirect_to_file_blocked.yaml
@@ -1,3 +1,4 @@
+# skip: redirect restrictions are an rshell-specific security feature
 skip_assert_against_bash: true
 description: Stderr redirection to a real file (not /dev/null) is still blocked.
 input:
diff --git a/tests/scenarios/shell/simple_command/basic/assignment_exit_status.yaml b/tests/scenarios/shell/simple_command/basic/assignment_exit_status.yaml
new file mode 100644
index 00000000..e887440a
--- /dev/null
+++ b/tests/scenarios/shell/simple_command/basic/assignment_exit_status.yaml
@@ -0,0 +1,10 @@
+description: Assignment-only command has exit status 0.
+input:
+  script: |+
+    a=1
+    echo $?
+expect:
+  stdout: |+
+    0
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/simple_command/basic/assignment_with_quotes.yaml b/tests/scenarios/shell/simple_command/basic/assignment_with_quotes.yaml
new file mode 100644
index 00000000..393d0331
--- /dev/null
+++ b/tests/scenarios/shell/simple_command/basic/assignment_with_quotes.yaml
@@ -0,0 +1,13 @@
+description: Assignment values can use quotes.
+input:
+  script: |+
+    a="hello world"
+    b='foo bar'
+    echo $a
+    echo $b
+expect:
+  stdout: |+
+    hello world
+    foo bar
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/simple_command/basic/command_not_found.yaml b/tests/scenarios/shell/simple_command/basic/command_not_found.yaml
new file mode 100644
index 00000000..635dbb48
--- /dev/null
+++ b/tests/scenarios/shell/simple_command/basic/command_not_found.yaml
@@ -0,0 +1,9 @@
+description: Unknown command produces not found error.
+input:
+  script: |+
+    nosuchcommand
+expect:
+  stdout: ""
+  stderr_contains:
+    - "command not found"
+  exit_code: 127
diff --git a/tests/scenarios/shell/simple_command/basic/multiple_assignments.yaml b/tests/scenarios/shell/simple_command/basic/multiple_assignments.yaml
new file mode 100644
index 00000000..d096d673
--- /dev/null
+++ b/tests/scenarios/shell/simple_command/basic/multiple_assignments.yaml
@@ -0,0 +1,10 @@
+description: Multiple variable assignments on one line.
+input:
+  script: |+
+    a=1 b=2
+    echo $a $b
+expect:
+  stdout: |+
+    1 2
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/simple_command/basic/redirect_between_tokens.yaml b/tests/scenarios/shell/simple_command/basic/redirect_between_tokens.yaml
new file mode 100644
index 00000000..1d1a8c83
--- /dev/null
+++ b/tests/scenarios/shell/simple_command/basic/redirect_between_tokens.yaml
@@ -0,0 +1,9 @@
+description: Redirection can appear between command arguments.
+input:
+  script: |+
+    echo 2>/dev/null foo
+expect:
+  stdout: |+
+    foo
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/simple_command/basic/single_assignment.yaml b/tests/scenarios/shell/simple_command/basic/single_assignment.yaml
new file mode 100644
index 00000000..cc3d656f
--- /dev/null
+++ b/tests/scenarios/shell/simple_command/basic/single_assignment.yaml
@@ -0,0 +1,10 @@
+description: Simple variable assignment and expansion.
+input:
+  script: |+
+    a=hello
+    echo $a
+expect:
+  stdout: |+
+    hello
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/until_clause/break_nested_for_in_until.yaml b/tests/scenarios/shell/until_clause/break_nested_for_in_until.yaml
new file mode 100644
index 00000000..94f04e9c
--- /dev/null
+++ b/tests/scenarios/shell/until_clause/break_nested_for_in_until.yaml
@@ -0,0 +1,14 @@
+# skip: until loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Break in for loop nested inside until loop is blocked.
+input:
+  script: |+
+    until false; do
+      for i in 1 2; do
+        break
+      done
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/until_clause/break_out_of_case.yaml b/tests/scenarios/shell/until_clause/break_out_of_case.yaml
new file mode 100644
index 00000000..ab6e50d4
--- /dev/null
+++ b/tests/scenarios/shell/until_clause/break_out_of_case.yaml
@@ -0,0 +1,14 @@
+# skip: until loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Break inside case statement within until loop is blocked.
+input:
+  script: |+
+    until false; do
+      case x in
+        x) break;;
+      esac
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/until_clause/break_out_of_else.yaml b/tests/scenarios/shell/until_clause/break_out_of_else.yaml
new file mode 100644
index 00000000..ec6a996c
--- /dev/null
+++ b/tests/scenarios/shell/until_clause/break_out_of_else.yaml
@@ -0,0 +1,16 @@
+# skip: until loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Break inside else branch within until loop is blocked.
+input:
+  script: |+
+    until false; do
+      if false; then
+        :
+      else
+        break
+      fi
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/until_clause/break_simple.yaml b/tests/scenarios/shell/until_clause/break_simple.yaml
new file mode 100644
index 00000000..37172362
--- /dev/null
+++ b/tests/scenarios/shell/until_clause/break_simple.yaml
@@ -0,0 +1,12 @@
+# skip: until loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Break in until loop is blocked because until loops are not supported.
+input:
+  script: |+
+    until false; do
+      break
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/until_clause/break_two_levels.yaml b/tests/scenarios/shell/until_clause/break_two_levels.yaml
new file mode 100644
index 00000000..cc7fed3b
--- /dev/null
+++ b/tests/scenarios/shell/until_clause/break_two_levels.yaml
@@ -0,0 +1,14 @@
+# skip: until loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Break 2 in nested until loops is blocked.
+input:
+  script: |+
+    until false; do
+      until false; do
+        break 2
+      done
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/until_clause/break_with_arg.yaml b/tests/scenarios/shell/until_clause/break_with_arg.yaml
new file mode 100644
index 00000000..50964199
--- /dev/null
+++ b/tests/scenarios/shell/until_clause/break_with_arg.yaml
@@ -0,0 +1,12 @@
+# skip: until loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Break with numeric argument in until loop is blocked.
+input:
+  script: |+
+    until false; do
+      break 1
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/until_clause/continue_nested_for_in_until.yaml b/tests/scenarios/shell/until_clause/continue_nested_for_in_until.yaml
new file mode 100644
index 00000000..902d1ea1
--- /dev/null
+++ b/tests/scenarios/shell/until_clause/continue_nested_for_in_until.yaml
@@ -0,0 +1,14 @@
+# skip: until loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Continue in for loop nested inside until loop is blocked.
+input:
+  script: |+
+    until false; do
+      for i in 1 2; do
+        continue
+      done
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/until_clause/continue_out_of_case.yaml b/tests/scenarios/shell/until_clause/continue_out_of_case.yaml
new file mode 100644
index 00000000..bdbe2ffc
--- /dev/null
+++ b/tests/scenarios/shell/until_clause/continue_out_of_case.yaml
@@ -0,0 +1,14 @@
+# skip: until loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Continue inside case statement within until loop is blocked.
+input:
+  script: |+
+    until false; do
+      case x in
+        x) continue;;
+      esac
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/until_clause/continue_out_of_else.yaml b/tests/scenarios/shell/until_clause/continue_out_of_else.yaml
new file mode 100644
index 00000000..dac18d11
--- /dev/null
+++ b/tests/scenarios/shell/until_clause/continue_out_of_else.yaml
@@ -0,0 +1,16 @@
+# skip: until loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Continue inside else branch within until loop is blocked.
+input:
+  script: |+
+    until false; do
+      if false; then
+        :
+      else
+        continue
+      fi
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/until_clause/continue_simple.yaml b/tests/scenarios/shell/until_clause/continue_simple.yaml
new file mode 100644
index 00000000..38b0f898
--- /dev/null
+++ b/tests/scenarios/shell/until_clause/continue_simple.yaml
@@ -0,0 +1,15 @@
+# skip: until loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Continue in until loop is blocked because until loops are not supported.
+input:
+  script: |+
+    i=0
+    until [ $i -gt 5 ]; do
+      i=$((i+1))
+      continue
+      echo unreachable
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/until_clause/continue_with_arg.yaml b/tests/scenarios/shell/until_clause/continue_with_arg.yaml
new file mode 100644
index 00000000..60b95f22
--- /dev/null
+++ b/tests/scenarios/shell/until_clause/continue_with_arg.yaml
@@ -0,0 +1,14 @@
+# skip: until loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Continue with numeric argument in until loop is blocked.
+input:
+  script: |+
+    i=0
+    until [ $i -gt 5 ]; do
+      i=$((i+1))
+      continue 1
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/until_clause/multiline_until.yaml b/tests/scenarios/shell/until_clause/multiline_until.yaml
index 31ad6ba1..dfafdc10 100644
--- a/tests/scenarios/shell/until_clause/multiline_until.yaml
+++ b/tests/scenarios/shell/until_clause/multiline_until.yaml
@@ -1,3 +1,4 @@
+# skip: until loops are intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Multi-line until loop is blocked.
 input:
diff --git a/tests/scenarios/shell/until_clause/until_with_brace.yaml b/tests/scenarios/shell/until_clause/until_with_brace.yaml
index 99b3c29b..97eb02cc 100644
--- a/tests/scenarios/shell/until_clause/until_with_brace.yaml
+++ b/tests/scenarios/shell/until_clause/until_with_brace.yaml
@@ -1,3 +1,4 @@
+# skip: until loops are intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Until loop with brace group body is blocked.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/all_params.yaml b/tests/scenarios/shell/var_expand/blocked_features/all_params.yaml
index b8ad5f53..f72bb9e2 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/all_params.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/all_params.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: All positional parameters $@ is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/all_params_star.yaml b/tests/scenarios/shell/var_expand/blocked_features/all_params_star.yaml
index bf4e9b4a..4241677c 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/all_params_star.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/all_params_star.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: All positional parameters $* is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/alternative.yaml b/tests/scenarios/shell/var_expand/blocked_features/alternative.yaml
index e7bf6b46..3f34d191 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/alternative.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/alternative.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Alternative value expansion ${var:+alt} is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/append.yaml b/tests/scenarios/shell/var_expand/blocked_features/append.yaml
index 74ddee11..0488fa50 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/append.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/append.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: The += append operator is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/arithmetic.yaml b/tests/scenarios/shell/var_expand/blocked_features/arithmetic.yaml
index 83b573d8..948d5325 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/arithmetic.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/arithmetic.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Arithmetic expansion is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/array.yaml b/tests/scenarios/shell/var_expand/blocked_features/array.yaml
index 577c67ae..4be90e86 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/array.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/array.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Array assignment is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/array_index.yaml b/tests/scenarios/shell/var_expand/blocked_features/array_index.yaml
index 9ca5cc91..a5fa627c 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/array_index.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/array_index.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Array indexing ${var[i]} is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/array_index_assign.yaml b/tests/scenarios/shell/var_expand/blocked_features/array_index_assign.yaml
index 16e5b663..7d320586 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/array_index_assign.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/array_index_assign.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Array index assignment arr[0]=value is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/assign_default.yaml b/tests/scenarios/shell/var_expand/blocked_features/assign_default.yaml
index e4978318..13bacfd9 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/assign_default.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/assign_default.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Assign default expansion ${var:=default} is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/case_conversion.yaml b/tests/scenarios/shell/var_expand/blocked_features/case_conversion.yaml
index fd07b720..8d8faf38 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/case_conversion.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/case_conversion.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Case conversion ${var^^} is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/command_sub.yaml b/tests/scenarios/shell/var_expand/blocked_features/command_sub.yaml
index 5ba0c494..d995cf18 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/command_sub.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/command_sub.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Command substitution is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/command_sub_backtick.yaml b/tests/scenarios/shell/var_expand/blocked_features/command_sub_backtick.yaml
index 5b211eea..bb096abe 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/command_sub_backtick.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/command_sub_backtick.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Backtick command substitution is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/default_value.yaml b/tests/scenarios/shell/var_expand/blocked_features/default_value.yaml
index b26cb28f..b6e9c45e 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/default_value.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/default_value.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Default value expansion ${var:-default} is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/error_unset.yaml b/tests/scenarios/shell/var_expand/blocked_features/error_unset.yaml
index a09eeca7..fb70bd82 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/error_unset.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/error_unset.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Error if unset expansion ${var:?message} is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/indirect.yaml b/tests/scenarios/shell/var_expand/blocked_features/indirect.yaml
index e930e969..79cf14ac 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/indirect.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/indirect.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Indirect expansion ${!var} is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/param_count.yaml b/tests/scenarios/shell/var_expand/blocked_features/param_count.yaml
index 82f74961..77763296 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/param_count.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/param_count.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Parameter count $# is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/positional_params.yaml b/tests/scenarios/shell/var_expand/blocked_features/positional_params.yaml
index 6ceb80aa..c6823e14 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/positional_params.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/positional_params.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Positional parameter $1 is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/prefix_list.yaml b/tests/scenarios/shell/var_expand/blocked_features/prefix_list.yaml
index af94c0fb..f58404ad 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/prefix_list.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/prefix_list.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Prefix variable listing ${!prefix*} is not supported (caught by indirect expansion check).
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/prefix_removal.yaml b/tests/scenarios/shell/var_expand/blocked_features/prefix_removal.yaml
index c691d64b..c47e5650 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/prefix_removal.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/prefix_removal.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Prefix removal ${var#pattern} is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/replace.yaml b/tests/scenarios/shell/var_expand/blocked_features/replace.yaml
index 1f0d00bd..bc33b852 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/replace.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/replace.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Pattern replacement ${var/pattern/replacement} is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/script_name.yaml b/tests/scenarios/shell/var_expand/blocked_features/script_name.yaml
index 52e3172c..e9d63e48 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/script_name.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/script_name.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Script name $0 is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/string_length.yaml b/tests/scenarios/shell/var_expand/blocked_features/string_length.yaml
index 57d544da..76c3b89d 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/string_length.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/string_length.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: String length expansion ${#var} is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/substring.yaml b/tests/scenarios/shell/var_expand/blocked_features/substring.yaml
index b1b51181..4338d798 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/substring.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/substring.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Substring expansion ${var:offset:length} is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_features/suffix_removal.yaml b/tests/scenarios/shell/var_expand/blocked_features/suffix_removal.yaml
index 2a1f59ca..cb1d9c37 100644
--- a/tests/scenarios/shell/var_expand/blocked_features/suffix_removal.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_features/suffix_removal.yaml
@@ -1,3 +1,4 @@
+# skip: variable expansion feature is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Suffix removal ${var%pattern} is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_variables/assignment.yaml b/tests/scenarios/shell/var_expand/blocked_variables/assignment.yaml
index cbfafde8..e03dbee1 100644
--- a/tests/scenarios/shell/var_expand/blocked_variables/assignment.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_variables/assignment.yaml
@@ -1,3 +1,4 @@
+# skip: special variable is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Variables without special meaning can be assigned and used as regular variables.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_variables/background_pid.yaml b/tests/scenarios/shell/var_expand/blocked_variables/background_pid.yaml
index 1bdd1964..2a469481 100644
--- a/tests/scenarios/shell/var_expand/blocked_variables/background_pid.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_variables/background_pid.yaml
@@ -1,3 +1,4 @@
+# skip: special variable is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: The $! variable is blocked.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_variables/euid.yaml b/tests/scenarios/shell/var_expand/blocked_variables/euid.yaml
index 5e716e15..f8acb2f9 100644
--- a/tests/scenarios/shell/var_expand/blocked_variables/euid.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_variables/euid.yaml
@@ -1,3 +1,4 @@
+# skip: special variable is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: The $EUID variable has no special meaning and expands to empty.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_variables/gid.yaml b/tests/scenarios/shell/var_expand/blocked_variables/gid.yaml
index 7b7a0c8c..6d25c036 100644
--- a/tests/scenarios/shell/var_expand/blocked_variables/gid.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_variables/gid.yaml
@@ -1,3 +1,4 @@
+# skip: special variable is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: The $GID variable has no special meaning and expands to empty.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_variables/in_braces.yaml b/tests/scenarios/shell/var_expand/blocked_variables/in_braces.yaml
index 493dd8f3..34eda626 100644
--- a/tests/scenarios/shell/var_expand/blocked_variables/in_braces.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_variables/in_braces.yaml
@@ -1,3 +1,4 @@
+# skip: special variable is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Variables without special meaning also expand to empty with brace syntax.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_variables/last_argument.yaml b/tests/scenarios/shell/var_expand/blocked_variables/last_argument.yaml
index a2eff5e8..4746ce04 100644
--- a/tests/scenarios/shell/var_expand/blocked_variables/last_argument.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_variables/last_argument.yaml
@@ -1,3 +1,4 @@
+# skip: special variable is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: The $_ variable (last argument) expands to empty.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_variables/lineno.yaml b/tests/scenarios/shell/var_expand/blocked_variables/lineno.yaml
index 79b083c4..89467edb 100644
--- a/tests/scenarios/shell/var_expand/blocked_variables/lineno.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_variables/lineno.yaml
@@ -1,3 +1,4 @@
+# skip: special variable is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: The $LINENO variable is not supported.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_variables/pid.yaml b/tests/scenarios/shell/var_expand/blocked_variables/pid.yaml
index 8512db4a..2daa0279 100644
--- a/tests/scenarios/shell/var_expand/blocked_variables/pid.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_variables/pid.yaml
@@ -1,3 +1,4 @@
+# skip: special variable is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: The $$ variable has no special meaning and expands to empty.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_variables/ppid.yaml b/tests/scenarios/shell/var_expand/blocked_variables/ppid.yaml
index c98fae0b..97fd54d2 100644
--- a/tests/scenarios/shell/var_expand/blocked_variables/ppid.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_variables/ppid.yaml
@@ -1,3 +1,4 @@
+# skip: special variable is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: The $PPID variable has no special meaning and expands to empty.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_variables/random.yaml b/tests/scenarios/shell/var_expand/blocked_variables/random.yaml
index 19037ce5..0489d841 100644
--- a/tests/scenarios/shell/var_expand/blocked_variables/random.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_variables/random.yaml
@@ -1,3 +1,4 @@
+# skip: special variable is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: The $RANDOM variable has no special meaning and expands to empty.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_variables/shell_options.yaml b/tests/scenarios/shell/var_expand/blocked_variables/shell_options.yaml
index 9132ea46..4aa040b1 100644
--- a/tests/scenarios/shell/var_expand/blocked_variables/shell_options.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_variables/shell_options.yaml
@@ -1,3 +1,4 @@
+# skip: special variable is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: The $- variable (shell options) expands to empty.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_variables/srandom.yaml b/tests/scenarios/shell/var_expand/blocked_variables/srandom.yaml
index 9aefa16d..47e48f50 100644
--- a/tests/scenarios/shell/var_expand/blocked_variables/srandom.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_variables/srandom.yaml
@@ -1,3 +1,4 @@
+# skip: special variable is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: The $SRANDOM variable has no special meaning and expands to empty.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_variables/stops_execution.yaml b/tests/scenarios/shell/var_expand/blocked_variables/stops_execution.yaml
index 9ed2255b..22ead782 100644
--- a/tests/scenarios/shell/var_expand/blocked_variables/stops_execution.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_variables/stops_execution.yaml
@@ -1,3 +1,4 @@
+# skip: special variable is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Variables without special meaning expand to empty but execution continues.
 input:
diff --git a/tests/scenarios/shell/var_expand/blocked_variables/uid.yaml b/tests/scenarios/shell/var_expand/blocked_variables/uid.yaml
index 95ad6fce..2bb5d340 100644
--- a/tests/scenarios/shell/var_expand/blocked_variables/uid.yaml
+++ b/tests/scenarios/shell/var_expand/blocked_variables/uid.yaml
@@ -1,3 +1,4 @@
+# skip: special variable is intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: The $UID variable has no special meaning and expands to empty.
 input:
diff --git a/tests/scenarios/shell/var_expand/quoting_and_escaping/backslash_in_dquotes_nonspecial.yaml b/tests/scenarios/shell/var_expand/quoting_and_escaping/backslash_in_dquotes_nonspecial.yaml
new file mode 100644
index 00000000..8a6941c8
--- /dev/null
+++ b/tests/scenarios/shell/var_expand/quoting_and_escaping/backslash_in_dquotes_nonspecial.yaml
@@ -0,0 +1,9 @@
+description: Backslash before non-special characters in double quotes is preserved.
+input:
+  script: |+
+    echo "\a\b\c"
+expect:
+  stdout: |+
+    \a\b\c
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/var_expand/quoting_and_escaping/backslash_special_chars.yaml b/tests/scenarios/shell/var_expand/quoting_and_escaping/backslash_special_chars.yaml
new file mode 100644
index 00000000..29b4d3bc
--- /dev/null
+++ b/tests/scenarios/shell/var_expand/quoting_and_escaping/backslash_special_chars.yaml
@@ -0,0 +1,9 @@
+description: Backslash escapes special characters in unquoted context.
+input:
+  script: |+
+    echo \&\(\)
+expect:
+  stdout: |+
+    &()
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/var_expand/quoting_and_escaping/double_quoted_not_globbed.yaml b/tests/scenarios/shell/var_expand/quoting_and_escaping/double_quoted_not_globbed.yaml
new file mode 100644
index 00000000..aa4ff84d
--- /dev/null
+++ b/tests/scenarios/shell/var_expand/quoting_and_escaping/double_quoted_not_globbed.yaml
@@ -0,0 +1,10 @@
+description: Glob characters in double-quoted expansion are not expanded.
+input:
+  script: |+
+    a="*"
+    echo "$a"
+expect:
+  stdout: |+
+    *
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/var_expand/quoting_and_escaping/double_quotes_backslash.yaml b/tests/scenarios/shell/var_expand/quoting_and_escaping/double_quotes_backslash.yaml
new file mode 100644
index 00000000..48b32318
--- /dev/null
+++ b/tests/scenarios/shell/var_expand/quoting_and_escaping/double_quotes_backslash.yaml
@@ -0,0 +1,9 @@
+description: Backslash in double quotes escapes $, backquote, double-quote, backslash, and newline.
+input:
+  script: |+
+    echo "a\\b" "a\$b"
+expect:
+  stdout: |+
+    a\b a$b
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/var_expand/quoting_and_escaping/double_quotes_expansion.yaml b/tests/scenarios/shell/var_expand/quoting_and_escaping/double_quotes_expansion.yaml
new file mode 100644
index 00000000..e7a28463
--- /dev/null
+++ b/tests/scenarios/shell/var_expand/quoting_and_escaping/double_quotes_expansion.yaml
@@ -0,0 +1,10 @@
+description: Variables are expanded inside double quotes.
+input:
+  script: |+
+    a=variable
+    echo "$a" "${a}"
+expect:
+  stdout: |+
+    variable variable
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/var_expand/quoting_and_escaping/double_quotes_multiline.yaml b/tests/scenarios/shell/var_expand/quoting_and_escaping/double_quotes_multiline.yaml
new file mode 100644
index 00000000..9e73cb6c
--- /dev/null
+++ b/tests/scenarios/shell/var_expand/quoting_and_escaping/double_quotes_multiline.yaml
@@ -0,0 +1,11 @@
+description: Double quotes preserve embedded newlines.
+input:
+  script: |+
+    echo "a
+    b"
+expect:
+  stdout: |+
+    a
+    b
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/var_expand/quoting_and_escaping/simplest_expansion.yaml b/tests/scenarios/shell/var_expand/quoting_and_escaping/simplest_expansion.yaml
new file mode 100644
index 00000000..59d58d8b
--- /dev/null
+++ b/tests/scenarios/shell/var_expand/quoting_and_escaping/simplest_expansion.yaml
@@ -0,0 +1,10 @@
+description: Basic variable expansion with $ and ${}.
+input:
+  script: |+
+    a=foo
+    echo $a ${a}
+expect:
+  stdout: |+
+    foo foo
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/var_expand/quoting_and_escaping/single_quotes.yaml b/tests/scenarios/shell/var_expand/quoting_and_escaping/single_quotes.yaml
new file mode 100644
index 00000000..498497e2
--- /dev/null
+++ b/tests/scenarios/shell/var_expand/quoting_and_escaping/single_quotes.yaml
@@ -0,0 +1,9 @@
+description: Single quotes preserve everything literally.
+input:
+  script: |+
+    echo 'abc' 'a\\b'
+expect:
+  stdout: |+
+    abc a\\b
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/var_expand/quoting_and_escaping/single_quotes_multiline.yaml b/tests/scenarios/shell/var_expand/quoting_and_escaping/single_quotes_multiline.yaml
new file mode 100644
index 00000000..678c34bc
--- /dev/null
+++ b/tests/scenarios/shell/var_expand/quoting_and_escaping/single_quotes_multiline.yaml
@@ -0,0 +1,11 @@
+description: Single quotes preserve embedded newlines.
+input:
+  script: |+
+    echo 'a
+    b'
+expect:
+  stdout: |+
+    a
+    b
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/var_expand/quoting_and_escaping/special_param_question.yaml b/tests/scenarios/shell/var_expand/quoting_and_escaping/special_param_question.yaml
new file mode 100644
index 00000000..fa779062
--- /dev/null
+++ b/tests/scenarios/shell/var_expand/quoting_and_escaping/special_param_question.yaml
@@ -0,0 +1,13 @@
+description: $? expands to the exit status of the last command.
+input:
+  script: |+
+    true
+    echo $?
+    false
+    echo $?
+expect:
+  stdout: |+
+    0
+    1
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/var_expand/quoting_and_escaping/unset_expands_empty.yaml b/tests/scenarios/shell/var_expand/quoting_and_escaping/unset_expands_empty.yaml
new file mode 100644
index 00000000..1414e031
--- /dev/null
+++ b/tests/scenarios/shell/var_expand/quoting_and_escaping/unset_expands_empty.yaml
@@ -0,0 +1,9 @@
+description: Unset variable expands to empty string.
+input:
+  script: |+
+    echo "[$unset_var]"
+expect:
+  stdout: |+
+    []
+  stderr: ""
+  exit_code: 0
diff --git a/tests/scenarios/shell/while_clause/break_nested_for_in_while.yaml b/tests/scenarios/shell/while_clause/break_nested_for_in_while.yaml
new file mode 100644
index 00000000..81d20f74
--- /dev/null
+++ b/tests/scenarios/shell/while_clause/break_nested_for_in_while.yaml
@@ -0,0 +1,14 @@
+# skip: while loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Break in for loop nested inside while loop is blocked.
+input:
+  script: |+
+    while true; do
+      for i in 1 2; do
+        break
+      done
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/while_clause/break_out_of_case.yaml b/tests/scenarios/shell/while_clause/break_out_of_case.yaml
new file mode 100644
index 00000000..d9be24ac
--- /dev/null
+++ b/tests/scenarios/shell/while_clause/break_out_of_case.yaml
@@ -0,0 +1,14 @@
+# skip: while loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Break inside case statement within while loop is blocked.
+input:
+  script: |+
+    while true; do
+      case x in
+        x) break;;
+      esac
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/while_clause/break_out_of_else.yaml b/tests/scenarios/shell/while_clause/break_out_of_else.yaml
new file mode 100644
index 00000000..6426b5f6
--- /dev/null
+++ b/tests/scenarios/shell/while_clause/break_out_of_else.yaml
@@ -0,0 +1,16 @@
+# skip: while loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Break inside else branch within while loop is blocked.
+input:
+  script: |+
+    while true; do
+      if false; then
+        :
+      else
+        break
+      fi
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/while_clause/break_simple.yaml b/tests/scenarios/shell/while_clause/break_simple.yaml
new file mode 100644
index 00000000..be1276b6
--- /dev/null
+++ b/tests/scenarios/shell/while_clause/break_simple.yaml
@@ -0,0 +1,12 @@
+# skip: while loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Break in while loop is blocked because while loops are not supported.
+input:
+  script: |+
+    while true; do
+      break
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/while_clause/break_two_levels.yaml b/tests/scenarios/shell/while_clause/break_two_levels.yaml
new file mode 100644
index 00000000..49c08754
--- /dev/null
+++ b/tests/scenarios/shell/while_clause/break_two_levels.yaml
@@ -0,0 +1,14 @@
+# skip: while loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Break 2 in nested while loops is blocked.
+input:
+  script: |+
+    while true; do
+      while true; do
+        break 2
+      done
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/while_clause/break_while_in_until.yaml b/tests/scenarios/shell/while_clause/break_while_in_until.yaml
new file mode 100644
index 00000000..a8a99d77
--- /dev/null
+++ b/tests/scenarios/shell/while_clause/break_while_in_until.yaml
@@ -0,0 +1,14 @@
+# skip: while loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Break in while loop nested inside until loop is blocked.
+input:
+  script: |+
+    until false; do
+      while true; do
+        break 2
+      done
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/while_clause/break_with_arg.yaml b/tests/scenarios/shell/while_clause/break_with_arg.yaml
new file mode 100644
index 00000000..a9b90e1a
--- /dev/null
+++ b/tests/scenarios/shell/while_clause/break_with_arg.yaml
@@ -0,0 +1,12 @@
+# skip: while loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Break with numeric argument in while loop is blocked.
+input:
+  script: |+
+    while true; do
+      break 1
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/while_clause/continue_nested_for_in_while.yaml b/tests/scenarios/shell/while_clause/continue_nested_for_in_while.yaml
new file mode 100644
index 00000000..404bab1f
--- /dev/null
+++ b/tests/scenarios/shell/while_clause/continue_nested_for_in_while.yaml
@@ -0,0 +1,14 @@
+# skip: while loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Continue in for loop nested inside while loop is blocked.
+input:
+  script: |+
+    while true; do
+      for i in 1 2; do
+        continue
+      done
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/while_clause/continue_out_of_case.yaml b/tests/scenarios/shell/while_clause/continue_out_of_case.yaml
new file mode 100644
index 00000000..1befb8be
--- /dev/null
+++ b/tests/scenarios/shell/while_clause/continue_out_of_case.yaml
@@ -0,0 +1,14 @@
+# skip: while loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Continue inside case statement within while loop is blocked.
+input:
+  script: |+
+    while true; do
+      case x in
+        x) continue;;
+      esac
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/while_clause/continue_out_of_else.yaml b/tests/scenarios/shell/while_clause/continue_out_of_else.yaml
new file mode 100644
index 00000000..b0b921ce
--- /dev/null
+++ b/tests/scenarios/shell/while_clause/continue_out_of_else.yaml
@@ -0,0 +1,16 @@
+# skip: while loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Continue inside else branch within while loop is blocked.
+input:
+  script: |+
+    while true; do
+      if false; then
+        :
+      else
+        continue
+      fi
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/while_clause/continue_simple.yaml b/tests/scenarios/shell/while_clause/continue_simple.yaml
new file mode 100644
index 00000000..bded0716
--- /dev/null
+++ b/tests/scenarios/shell/while_clause/continue_simple.yaml
@@ -0,0 +1,15 @@
+# skip: while loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Continue in while loop is blocked because while loops are not supported.
+input:
+  script: |+
+    i=0
+    while [ $i -lt 5 ]; do
+      i=$((i+1))
+      continue
+      echo unreachable
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/while_clause/continue_while_in_until.yaml b/tests/scenarios/shell/while_clause/continue_while_in_until.yaml
new file mode 100644
index 00000000..aa4a0a68
--- /dev/null
+++ b/tests/scenarios/shell/while_clause/continue_while_in_until.yaml
@@ -0,0 +1,14 @@
+# skip: while loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Continue in while loop nested inside until loop is blocked.
+input:
+  script: |+
+    until false; do
+      while true; do
+        continue 2
+      done
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/while_clause/continue_with_arg.yaml b/tests/scenarios/shell/while_clause/continue_with_arg.yaml
new file mode 100644
index 00000000..488aac3e
--- /dev/null
+++ b/tests/scenarios/shell/while_clause/continue_with_arg.yaml
@@ -0,0 +1,14 @@
+# skip: while loops are intentionally blocked in the restricted shell
+skip_assert_against_bash: true
+description: Continue with numeric argument in while loop is blocked.
+input:
+  script: |+
+    i=0
+    while [ $i -lt 5 ]; do
+      i=$((i+1))
+      continue 1
+    done
+expect:
+  stdout: ""
+  stderr: "while/until loops are not supported\n"
+  exit_code: 2
diff --git a/tests/scenarios/shell/while_clause/multiline_while.yaml b/tests/scenarios/shell/while_clause/multiline_while.yaml
index 129a3987..a3af6646 100644
--- a/tests/scenarios/shell/while_clause/multiline_while.yaml
+++ b/tests/scenarios/shell/while_clause/multiline_while.yaml
@@ -1,3 +1,4 @@
+# skip: while loops are intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: Multi-line while loop is blocked.
 input:
diff --git a/tests/scenarios/shell/while_clause/while_with_brace.yaml b/tests/scenarios/shell/while_clause/while_with_brace.yaml
index 8edeb4b5..a0363e6b 100644
--- a/tests/scenarios/shell/while_clause/while_with_brace.yaml
+++ b/tests/scenarios/shell/while_clause/while_with_brace.yaml
@@ -1,3 +1,4 @@
+# skip: while loops are intentionally blocked in the restricted shell
 skip_assert_against_bash: true
 description: While loop with brace group body is blocked.
 input: