ai-slop-cleaner

Lock behavior with targeted regression tests before editing, then execute cleanup one smell category at a time: dead code deletion, duplicate removal, naming and error handling cleanup, and test reinforcement. Re-run verification after each pass and keep the diff minimal and scoped.

Use a bounded deslop workflow: create an explicit cleanup plan, inventory fallback-like code, categorize smells, then run ordered passes with quality gates for lint, typecheck, and tests. Avoid bundling unrelated refactors into one large edit set.

Yes. The skill accepts a file list scope, such as changed files from an automated coding session, and keeps the entire cleanup pass strictly bounded to that list instead of touching the whole feature area.

Each fallback-like finding is classified as masking fallback slop or grounded compatibility/fail-safe fallback. Masking fallbacks are removed or repaired at the root cause, while grounded fallbacks with documentation and regression tests for both paths may be preserved.

Broad, ambiguous, cross-layer, or architectural fallback-like code should be escalated to a consensus planning workflow before edits. When already inside such a workflow, findings are recorded and attached to the active plan instead of spawning a nested one.