diff --git a/packages/guardrails/profile/commands/blog.md b/packages/guardrails/profile/commands/blog.md
new file mode 100644
index 000000000000..dfb283e963f7
--- /dev/null
+++ b/packages/guardrails/profile/commands/blog.md
@@ -0,0 +1,26 @@
+---
+description: Generate a blog article from provided topic or context.
+agent: implement
+---
+
+Write a technical blog post based on the provided topic.
+
+1. Identify the topic from $ARGUMENTS or the current conversation context.
+2. Research the topic within the codebase for concrete examples and code snippets.
+3. Structure the post with a clear narrative arc.
+4. Include code examples that are accurate and runnable.
+5. Match the project's technical voice and terminology.
+
+Required structure:
+- Title
+- Introduction (hook + what the reader will learn)
+- Main sections (2-4 sections with headers, explanations, and code examples)
+- Conclusion (summary + next steps or call to action)
+
+Guidelines:
+- Keep paragraphs short — 3-4 sentences maximum.
+- Code examples must be syntax-highlighted with the correct language tag.
+- Avoid jargon without explanation.
+- Target length: 800-1500 words.
+
+$ARGUMENTS
diff --git a/packages/guardrails/profile/commands/code-review.md b/packages/guardrails/profile/commands/code-review.md
new file mode 100644
index 000000000000..d7ef697e9938
--- /dev/null
+++ b/packages/guardrails/profile/commands/code-review.md
@@ -0,0 +1,23 @@
+---
+description: Detailed code review for security, quality, and best practices.
+agent: code-reviewer
+subtask: true
+---
+
+Review code for bugs, logic errors, security vulnerabilities, code quality, and adherence to project conventions.
+
+1. Identify the review scope from git diff or $ARGUMENTS.
+2. Check for correctness — logic errors, off-by-one, null handling, race conditions.
+3. Check for security — injection, XSS, CSRF, secrets exposure, auth/authz gaps.
+4. Check for quality — naming, duplication, complexity, test coverage gaps.
+5. Check for performance — unnecessary allocations, N+1 queries, missing indexes.
+6. Verify adherence to project conventions and existing patterns.
+
+Required sections:
+
+- Findings (grouped by severity: Critical, High, Medium, Low)
+- Security concerns
+- Performance issues
+- Recommendations
+
+Default scope is the current uncommitted work unless `$ARGUMENTS` narrows it.
diff --git a/packages/guardrails/profile/commands/explain-project.md b/packages/guardrails/profile/commands/explain-project.md
new file mode 100644
index 000000000000..88586b5675bc
--- /dev/null
+++ b/packages/guardrails/profile/commands/explain-project.md
@@ -0,0 +1,24 @@
+---
+description: Explain the high-level architecture of the current project.
+agent: planner
+---
+
+Analyze the project and explain its architecture.
+
+1. Scan the project root for configuration files, entry points, and directory structure.
+2. Identify main components, modules, and their responsibilities.
+3. Trace data flow between components (API boundaries, state management, persistence).
+4. Detect architecture patterns in use (MVC, hexagonal, event-driven, monorepo, etc.).
+5. Map key dependencies and their roles.
+6. Use multiple subagents if the project spans many packages or domains.
+
+Required output:
+- Architecture overview (one paragraph)
+- Key files and directories (with purpose)
+- Dependency graph (text or mermaid)
+- Design patterns identified
+- Data flow summary
+
+$ARGUMENTS narrows scope to a specific area if provided.
+
+$ARGUMENTS
diff --git a/packages/guardrails/profile/commands/gemini.md b/packages/guardrails/profile/commands/gemini.md
new file mode 100644
index 000000000000..b67b671f7f9c
--- /dev/null
+++ b/packages/guardrails/profile/commands/gemini.md
@@ -0,0 +1,20 @@
+---
+description: Web search using Gemini CLI for research.
+agent: implement
+---
+
+Use the `gemini` CLI tool to perform web search and return synthesized results.
+
+1. Take the search query from $ARGUMENTS.
+2. Run `gemini` CLI with the provided query.
+3. Parse and organize the search results.
+4. Synthesize findings into actionable information relevant to the current task.
+5. Cite sources where applicable.
+
+Required output:
+- Search query used
+- Key findings (bulleted, ranked by relevance)
+- Source references
+- Actionable recommendations based on findings
+
+$ARGUMENTS
diff --git a/packages/guardrails/profile/commands/plan-to-checklist.md b/packages/guardrails/profile/commands/plan-to-checklist.md
new file mode 100644
index 000000000000..7d2a5ced4d5b
--- /dev/null
+++ b/packages/guardrails/profile/commands/plan-to-checklist.md
@@ -0,0 +1,22 @@
+---
+description: Convert a plan into a checklist document.
+agent: planner
+---
+
+Take the provided plan and convert it into a structured checklist document.
+
+1. Read the plan from $ARGUMENTS or the current conversation context.
+2. Break the plan into phases or priority groups.
+3. Convert each item into a checkbox entry with a clear, actionable description.
+4. Add dependencies between items where they exist.
+5. Include acceptance criteria for each item so completion is verifiable.
+6. Output the checklist as a markdown document.
+
+Required structure:
+- Document title (from $ARGUMENTS or inferred from the plan)
+- Phases or priority groups with headers
+- Checkbox items (`- [ ]`) under each group
+- Dependencies noted inline (e.g., "depends on: item X")
+- Acceptance criteria per item (indented under the checkbox)
+
+$ARGUMENTS
diff --git a/packages/guardrails/profile/commands/test-report.md b/packages/guardrails/profile/commands/test-report.md
new file mode 100644
index 000000000000..fe8b5d118efd
--- /dev/null
+++ b/packages/guardrails/profile/commands/test-report.md
@@ -0,0 +1,22 @@
+---
+description: Run tests and generate a report saved to log directory.
+agent: implement
+---
+
+Run the project test suite and generate a structured report.
+
+1. Detect the project test framework and runner (jest, vitest, pytest, go test, etc.).
+2. Run the full test suite or the scope specified in $ARGUMENTS.
+3. Collect results including pass/fail/skip counts, duration, and coverage if available.
+4. Save the report to `./log/test/` with a timestamped filename (e.g., `test-report-2026-04-06T12-00-00.md`).
+5. If tests fail, include failure details with file and line references.
+
+Required output in the saved report:
+- Test command executed
+- Pass / fail / skip counts
+- Coverage summary (if available)
+- Failed test details (name, file, error message)
+- Total duration
+- Timestamp
+
+$ARGUMENTS
diff --git a/packages/guardrails/profile/commands/ui-skills.md b/packages/guardrails/profile/commands/ui-skills.md
new file mode 100644
index 000000000000..fe39d653f1dc
--- /dev/null
+++ b/packages/guardrails/profile/commands/ui-skills.md
@@ -0,0 +1,24 @@
+---
+description: UI component guidelines and constraints for building interfaces.
+agent: implement
+---
+
+Provide opinionated UI constraints and guidelines for the current project.
+
+1. Detect the UI framework and styling approach in use (React, Vue, Svelte, Tailwind, etc.).
+2. Apply Tailwind CSS best practices — utility-first, consistent spacing scale, design tokens.
+3. Enforce accessibility requirements — ARIA roles, keyboard navigation, color contrast, focus management.
+4. Define animation guidelines — prefer CSS transitions, respect prefers-reduced-motion, keep durations under 300ms.
+5. Establish responsive layout patterns — mobile-first breakpoints, fluid typography, container queries.
+6. Guide component composition — single responsibility, prop drilling limits, slot/children patterns.
+
+Guidelines:
+- No inline styles — use Tailwind utilities or CSS modules.
+- Interactive elements must have visible focus indicators.
+- Images require alt text; decorative images use `alt=""`.
+- Modals and dialogs must trap focus and support Escape to close.
+- Color must not be the only indicator of state (use icons or text).
+
+Apply constraints to the current context or the scope specified in $ARGUMENTS.
+
+$ARGUMENTS