| name | looping-quality |
| description | Run scanning-quality, fix findings, repeat until clean or the iteration limit is hit. |
| user-invocable | true |
| allowed-tools | Skill, Task, Read, Grep, Glob, Bash(pnpm run check:*), Bash(pnpm run test:*), Bash(pnpm test:*), Bash(pnpm run build:*), Bash(git status:*), Bash(git diff:*), Bash(git add:*), Bash(git commit:*) |
| model | claude-sonnet-4-6 |
| metadata | {"internal":true} |
looping-quality
A thin loop counter over the scanning-quality
primitive. scanning-quality is one pass — fan out finders, dedup, verify,
produce an A-F report. This skill wraps it in an iterate-fix-recheck loop: scan,
fix the findings, scan again, until the report is clean or the iteration cap is
hit. All the scanning logic lives in scanning-quality; this skill only adds the
counter and the fix-and-recheck cadence.
Interactive only — this skill makes code changes and commits. Do not use as
an automated pipeline gate (that's what a single scanning-quality report is
for).
Process
Track an iteration counter N, starting at 1, capped at MAX_ITERATIONS = 5.
- Scan. Run the
scanning-quality skill (all scan types). It returns the
A-F report + findings.
- Done check. If zero findings → success; report the clean pass and stop.
- Fix. Spawn the
refactor-cleaner agent (see agents/refactor-cleaner.md)
to fix the findings, grouped by category. Honor CLAUDE.md's pre-action
protocol: dead code first, then structural changes, ≤5 files per phase.
- Verify. Run verify-build (see
_shared/verify-build.md) and the test
suite after fixes to confirm nothing broke.
- Commit.
fix: resolve quality scan issues (iteration N).
- Loop. Increment
N. If N > MAX_ITERATIONS, stop and report remaining
findings. Otherwise go to step 1.
Rules
- Fix every finding, not just the easy ones.
- One commit per iteration; the iteration number is in the commit subject so the
trend is visible in
git log.
- Run tests after each fix batch — a fix that breaks the build is not a fix.
- The heavy scanning work is delegated to
scanning-quality (which pins opus);
this skill just orchestrates the loop, so it runs on a lighter model.
- Report the final state: iterations run, findings fixed, anything still open at
the cap.
