codebase-cleaner

name	codebase-cleaner
description	Cleanup worker for deleting AI slop, consolidating duplicates, and fixing imports. Never breaks tests.

Codebase Cleaner

NOTE: Startup is handled by worker-base. This skill defines the WORK PROCEDURE for cleanup tasks.

When to Use This Skill

Use for features that:

• Delete files or directories from the repository
• Consolidate duplicate modules into a single canonical location
• Remove shim/re-export directories
• Clean up documentation padding
• Update imports after moving or deleting files

NOT for: writing new features, TDD workflows, or fixing broken logic (use python-developer for those).

Work Procedure

1. Baseline Verification

Before touching anything, run the test suite and record the baseline:

python -m pytest tests/unit/ tests/integration/ tests/performance -n 4 --ignore=tests/poc -q

Record: X passed, Y failed. If Y > 0, stop and return to orchestrator — do not proceed with cleanup against a broken baseline.

2. Investigate Before Deleting

For each file or directory you plan to delete:

git ls-files <path>                                  # confirm it exists in git
rg 'from reveng\.<module>|import reveng\.<module>' src/ tests/ --type py  # find all importers

If importers exist, update them BEFORE deleting the source. Never delete a file that has live importers without first updating them.

3. The Cleanup Loop (repeat for each batch)

a) Update imports (if any) in dependent files
b) Run tests: python -m pytest tests/unit/ tests/integration/ tests/performance -n 4 --ignore=tests/poc -q
   → Must pass with 0 failures before proceeding
c) git rm <files>  (use git rm, not plain rm, so deletions are staged)
d) Run tests again to confirm nothing broke
e) git commit -m "cleanup: <description of what was removed>"

Keep commits small and frequent. One commit per logical batch (e.g., "cleanup: delete all claude.md files").

4. Import Update Pattern

When moving a module from A to B, update all importers:

# Before: from reveng.tools.ai import SomeClass
# After:  from reveng.agents.ai import SomeClass

Use rg to find every import site, then edit each file. After updating, run tests before deleting the original.

5. Directory Deletion

Only delete a directory when ALL its files have been removed:

git ls-files <dir/>  # must return empty before directory deletion
git rm -r <dir/>     # removes any remaining tracked files

6. Verification

After completing all work for the feature:

git ls-files | rg '<pattern>'  # confirm deletions took effect
python -m pytest tests/unit/ tests/integration/ tests/performance -n 4 --ignore=tests/poc -q
git status  # must be clean after commit

CRITICAL: Lint and typecheck scope for cleanup features

• Run flake8 and mypy ONLY on the files you touched — NOT repo-wide.
• Repo-wide flake8 src/reveng/ and mypy src/reveng/ have pre-existing failures unrelated to cleanup work.
• Pre-existing lint/mypy failures in files you did NOT touch are NOT your problem and must NOT block your commit.
• If the scoped lint on your touched files passes AND pytest passes with 0 failures, commit and proceed.
• Only block if you introduced NEW failures in files you touched.

7. Commit and Report

Single clean commit per feature (or small sequential commits). The handoff must include:

• Exact git ls-files commands used for verification
• Before/after file counts
• Test output confirming 0 failures

Example Handoff

{
  "salientSummary": "Deleted all 113 claude.md files with git rm. Test suite confirmed 0 regressions (307 passed, 0 failed). git ls-files | rg 'claude.md' | wc -l = 0.",
  "whatWasImplemented": "Ran `git ls-files | rg 'claude\\.md'` to enumerate all 113 AI context files, then `git rm` each in batches. Ran baseline tests before and after — no failures either time. Committed as a single cleanup commit.",
  "whatWasLeftUndone": "",
  "verification": {
    "commandsRun": [
      {
        "command": "git ls-files | rg 'claude\\.md' | wc -l",
        "exitCode": 0,
        "observation": "Confirmed 113 claude.md files in git index before deletion"
      },
      {
        "command": "python -m pytest tests/unit/ tests/integration/ tests/performance -n 4 --ignore=tests/poc -q",
        "exitCode": 0,
        "observation": "307 passed, 0 failed, 93 skipped — baseline confirmed clean before deletion"
      },
      
         
         
         
      
      
         
         
         
      
      
         
         
         
      
    
  
   
     
     
  
   

When to Return to Orchestrator

• Test baseline is already red before your cleanup starts
• An import you need to update is complex (circular imports, dynamic imports, __all__ re-exports that span multiple levels)
• A directory you were asked to delete actually contains non-shim code that's still needed
• After a deletion, tests fail and you cannot determine which import is broken