| name | qa-audit |
| description | Audit one bash feature against the QA rubric and write its section into docs/testing/interactive-parity-audit.md |
QA AUDIT. ONE FEATURE PER RUN. Ref: @.claude/rules/qa-rubric.md (D10 = TEMPLATE).
文言:審一feature——尋測試、核15失敗軸、核6模式、核oracle、核已知患,依D10模板寫一節;勿寫測、勿改碼。
INPUT: $ARGUMENTS = feature (e.g. "pipes", "if/elif/else", "command substitution").
STEPS
- RESOLVE → TEST FILES. grep
src/**/*.Tests/** + PsBash.psm1 (if runtime). List file:test. No prose.
- FAILURE-SURFACE: each of D3's 15 axes → YES/NO/PARTIAL (justify skip, 1 line).
- MODE: each of D4's M1..M6 → YES/NO/PARTIAL.
- ORACLE: grep tests for differential harness (Phase 0 fixture). None → NO + why (D1 exceptions).
- KNOWN BUGS: grep
docs/solutions/ + Dart (tags: bug-fix / feature). List file:line or Dart ID.
- WRITE one section, D10 template EXACTLY → append
docs/testing/interactive-parity-audit.md. No prose outside template. Priority gaps: top 3, by user impact.
- SUMMARY to user: feature, gap count, P1 gap (1 line), link.
BARS
One new section, ≤60 md lines, template-conformant. grep "### FEATURE:" … | wc -l == audited count. Consistent columns.
DO NOT
No writing tests. No fixing bugs. No editing other sections (only refresh one feature).