Codex AI Slop Cleaner

Clean AI-generated slop without changing intended behavior. This is the native Codex adaptation of OMX $ai-slop-cleaner: regression checks first, bounded scope, smell-by-smell cleanup, fallback classification, and evidence-dense reporting.

Contract

• Preserve behavior unless the user explicitly asks to change it.
• Keep scope bounded to requested files or the current workflow's changed files.
• Do not perform broad rewrites, architecture pivots, or dependency additions without approval.
• Protect user work and unrelated dirty files.
• Verify after each meaningful cleanup pass.

Workflow

1. Lock behavior with targeted tests, smoke checks, or visual evidence.
2. Make a cleanup plan naming smells and file scope.
3. Classify fallback-like code before editing:
   • Masking slop: swallowed errors, silent defaults, untested alternate paths, broad bypasses.
   • Grounded fallback: scoped compatibility/fail-safe boundary with rationale and tests.
4. Execute one smell category at a time: dead code, duplication, needless wrappers, naming/error handling, test reinforcement, or UI/design slop.
5. Verify after each pass.
6. Report changed files, simplifications, tests, fallback decisions, and remaining risks.

Slop Signals

• "temporary" code that became permanent
• fallback branches that hide errors
• generic wrappers around one call site
• repeated content scaffolding or duplicated helpers
• comments that narrate obvious code
• over-engineered option objects or config layers
• default AI palettes, gratuitous gradients, emoji badges, card grids, and filler UI copy without product reason

Output Shape

• Scope
• Behavior lock
• Cleanup plan
• Passes completed
• Verification
• Changed files
• Fallback review
• Remaining risks

References

Read references/slop-checklist.html for smell categories and cleanup gates.