| name | slop-sweep |
| description | Structured scan for AI-slop code patterns across backend + frontend. Use when the user says "run /slop-sweep", asks to review recent diffs for AI-generated cruft, or wants to sweep the tree for banner comments / as-unknown-as casts / stacked any-suppressions / mock-data imports in production. |
| tools | ["Read","Bash","Glob"] |
Slop Sweep
Seven-check sweep for the low-signal patterns that show up in
AI-generated diffs. Output is a prioritized findings list — high-severity
first (guardrail violations), low-severity last (chatter).
How to run
python .claude/skills/slop-sweep/check.py
Checks
| # | Severity | Pattern | Why it matters |
|---|
| 1 | HIGH | Underscore-prefixed Depends(...) parameters in backend/app/routes/ | Silences linters on audit/http_request — CLAUDE.md guardrail #2. |
| 2 | HIGH | mockData imports in non-test frontend files | CLAUDE.md guardrail #6 — mock data must never ship. |
| 3 | MED | as unknown as X casts in TypeScript | Usually a shortcut past real typing; CLAUDE.md guardrail #7. |
| 4 | MED | @typescript-eslint/no-explicit-any stacked ≥2× in one file | Three or more identical casts is a refactor obligation, not a lint exception (guardrail #7). |
| 5 | MED | Hardcoded hex colors (#abc123) outside tailwind.config.ts | Pablo uses the design-system palette; raw hex is usually slop. |
| 6 | LOW | Banner comments (# ==== FOO ====, // ==== FOO ====) | No semantic value; AI padding. |
| 7 | LOW | // TODO: consider... / # TODO: consider... without an issue link | Unactionable chatter. |
Output
Grouped by severity, one finding per line (file:line pattern snippet),
followed by a totals row. Exit code 1 if any HIGH finding, otherwise 0
(MED/LOW are reported but don't fail the command — they're for review).
Scope
- Backend:
backend/app/**/*.py
- Frontend:
frontend/src/**/*.{ts,tsx} and frontend/app/**/*.{ts,tsx}
- Ignores
node_modules/, .next/, dist/, build/, __pycache__/,
and any file under a test/ or tests/ directory (for the mockData
check only).
