diff --git a/.github/workflows/autoloop.md b/.github/workflows/autoloop.md
index 0c2d9c3e..6b9dfdd8 100644
--- a/.github/workflows/autoloop.md
+++ b/.github/workflows/autoloop.md
@@ -293,7 +293,7 @@ A template program file is installed at `.autoloop/programs/example.md`. **Progr
 At the start of every run, check each program file for this sentinel. For any program where it is present:
 
 1. **Skip that program — do not run any iterations for it.**
-2. If no setup issue exists for that program, create one titled `[Autoloop: {program-name}] Action required: configure your program`.
+2. If no setup issue exists for that program, create one titled `[Autoloop: {program-name}] Action required: configure your program` (supply this exact string — do NOT prepend `[Autoloop] `, the `create-issue` safe-output adds that automatically).
 
 ## Branching Model
 
@@ -570,7 +570,7 @@ There are no separate "steering" or "experiment log" issues — they have all be
 
 If `selected_issue` is `null` in `/tmp/gh-aw/autoloop.json`, the program is file-based **and** has no program issue yet. On the first run, create one with `create-issue`:
 
-- **Title**: `[Autoloop: {program-name}]`
+- **Title**: `[Autoloop: {program-name}]` — supply this exact string as the title. **Do NOT prepend `[Autoloop] `** yourself; the `create-issue` safe-output's `title-prefix` config adds that automatically. The final issue title will be `[Autoloop] [Autoloop: {program-name}]` on GitHub, and the scheduler recognizes that form when matching a program issue back to its file-based program.
 - **Labels**: `autoloop-program`, `automation`, `autoloop`
 - **Body**: the contents of the program file (so humans can read the goal/target/evaluation directly on the issue), prefixed with `🤖 *Autoloop program issue for `{program-name}`. The program definition below is mirrored from [`{selected_file}`]({link-to-file}). Edit the file to update the definition; comment on this issue to steer the agent.*`
 
diff --git a/.github/workflows/scripts/autoloop_scheduler.py b/.github/workflows/scripts/autoloop_scheduler.py
index 64395293..925829d1 100644
--- a/.github/workflows/scripts/autoloop_scheduler.py
+++ b/.github/workflows/scripts/autoloop_scheduler.py
@@ -14,6 +14,57 @@
 autoloop_dir = ".autoloop/programs"
 template_file = os.path.join(autoloop_dir, "example.md")
 
+# Regex matching the canonical program-issue title, tolerating the
+# `[Autoloop] ` safe-outputs prefix that the `create-issue` machinery
+# auto-prepends. Both `[Autoloop: name]` (raw, before prefix is added) and
+# `[Autoloop] [Autoloop: name]` (after prefix is applied) are recognised so
+# the scheduler matches a program issue back to its file-based program
+# regardless of whether the agent or the safe-outputs layer added the
+# outer marker. The match is case-insensitive for robustness.
+_AUTOLOOP_PREFIX_RE = re.compile(r'^\[Autoloop\]\s*', re.IGNORECASE)
+_AUTOLOOP_NAME_RE = re.compile(r'^\[Autoloop:\s*([^\]]+?)\s*\]\s*$', re.IGNORECASE)
+
+def extract_program_name_from_issue_title(title):
+    """Return the program-name embedded in a canonical program-issue title.
+
+    Accepts titles of the form `[Autoloop: name]` and tolerates any number
+    of leading `[Autoloop] ` prefixes (the safe-outputs prefix can collide
+    with an agent-supplied `[Autoloop]` marker, producing doubly-prefixed
+    titles like `[Autoloop] [Autoloop: name]`). Returns ``None`` when the
+    title does not match the canonical pattern — callers should then fall
+    back to slugification for human-authored issue titles.
+    """
+    if not title:
+        return None
+    s = title.strip()
+    while _AUTOLOOP_PREFIX_RE.match(s):
+        s = _AUTOLOOP_PREFIX_RE.sub('', s, count=1)
+    m = _AUTOLOOP_NAME_RE.match(s)
+    if m:
+        return m.group(1).strip()
+    return None
+
+def slugify(title):
+    """Slugify an issue title to a program name.
+
+    Defensively strips any leading `[Autoloop] ` markers (collapses
+    repeated prefixes) and a `[Autoloop: name]` wrapper before
+    slugifying, so doubly-prefixed titles authored under an old or buggy
+    prompt still collapse to the same slug as the canonical name. This
+    makes the scheduler self-healing: even if Fix 1 (prompt clarification)
+    regresses, this normalisation keeps file-based and issue-based
+    discovery from forking into two programs.
+    """
+    s = (title or "").strip()
+    while _AUTOLOOP_PREFIX_RE.match(s):
+        s = _AUTOLOOP_PREFIX_RE.sub('', s, count=1)
+    m = re.match(r'^\[Autoloop:\s*([^\]]+?)\s*\]\s*', s, re.IGNORECASE)
+    if m:
+        s = m.group(1)
+    slug = re.sub(r'[^a-z0-9]+', '-', s.lower()).strip('-')
+    slug = re.sub(r'-+', '-', slug)
+    return slug
+
 # Read program state from repo-memory (persistent git-backed storage)
 github_token = os.environ.get("GITHUB_TOKEN", "")
 repo = os.environ.get("GITHUB_REPOSITORY", "")
@@ -164,17 +215,20 @@ def read_program_state(program_name):
             known_file_program_names.add(os.path.basename(os.path.dirname(pf)))
         else:
             known_file_program_names.add(os.path.splitext(os.path.basename(pf))[0])
-    file_program_issue_pattern = re.compile(r'^\s*\[Autoloop:\s*([^\]]+?)\s*\]\s*$')
     consumed_issue_numbers = set()
     for issue in issues:
         if issue.get("pull_request"):
             continue
         title = issue.get("title") or ""
-        m = file_program_issue_pattern.match(title)
-        if m and m.group(1) in known_file_program_names:
-            file_program_issues[m.group(1)] = issue["number"]
+        # extract_program_name_from_issue_title tolerates the doubly-prefixed
+        # `[Autoloop] [Autoloop: name]` form produced when the safe-outputs
+        # `title-prefix` collides with an agent-supplied marker, so existing
+        # in-the-wild issues still merge with their file-based program here.
+        extracted = extract_program_name_from_issue_title(title)
+        if extracted and extracted in known_file_program_names:
+            file_program_issues[extracted] = issue["number"]
             consumed_issue_numbers.add(issue["number"])
-            print(f"  Found program issue for file-based program '{m.group(1)}': #{issue['number']}")
+            print(f"  Found program issue for file-based program '{extracted}': #{issue['number']}")
 
     # Second pass: any remaining autoloop-program issue is an issue-based program.
     for issue in issues:
@@ -185,9 +239,12 @@ def read_program_state(program_name):
         body = issue.get("body") or ""
         title = issue.get("title") or ""
         number = issue["number"]
-        # Derive program name from issue title: slugify to lowercase with hyphens
-        slug = re.sub(r'[^a-z0-9]+', '-', title.lower()).strip('-')
-        slug = re.sub(r'-+', '-', slug)  # collapse consecutive hyphens
+        # Derive program name from issue title via the defensive slugify
+        # (strips known `[Autoloop]`/`[Autoloop: name]` prefixes before
+        # slugifying, so a stray doubly-prefixed title that didn't match a
+        # known file-based program still produces a clean slug rather than
+        # a `autoloop-autoloop-...` chimera).
+        slug = slugify(title)
         if not slug:
             slug = f"issue-{number}"
         # Avoid slug collisions: if another issue already claimed this slug, append issue number
diff --git a/.gitignore b/.gitignore
index 354816e7..0b8c8525 100644
--- a/.gitignore
+++ b/.gitignore
@@ -5,3 +5,5 @@ package-lock.json
 *.tgz
 playground/benchmarks/
 playground/dist/
+__pycache__/
+*.pyc