Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Review a Spec Kitty work package through the runtime review surface and approve or reject with structured feedback.
.claude/skills/priivacy-ai-spk-run-review-wp/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -60% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -33% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -69% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -67% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -75% | 0% |
Use this skill when the user asks to review a WP, approve/reject work, or operate the review workflow surface.
spk-run-implement-review continue the loop.Capture the approve/reject decision through the deterministic event-log seam described in spk-run-verdict-capture — the review_result event in status.events.jsonl is the sole authority (the review-cycle-N.md render is non-authoritative). In practice: spec-kitty agent tasks move-task <WP> --to approved to approve, or --to planned --review-feedback-file <f> to reject. Never hand-edit the .md render to change a verdict.
For detailed review command behavior, use spec-kitty-runtime-review when available.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-07 | fail→pass | 11,446 | 2,857 | -75% | 1 | 1 | 0% | 1,703 | 686 | -60% | 0 | 0 | — |
case-01 | fail→fail | 13,706 | 5,463 | -60% | 1 | 1 | 0% | 349 | 466 | +34% | 0 | 0 | — |
case-16 | fail→pass | 9,277 | 3,967 | -57% | 1 | 1 | 0% | 1,371 | 925 | -33% | 0 | 0 | — |
case-02 | fail→fail | 3,550 | 6,565 | +85% | 1 | 1 | 0% | 445 | 556 | +25% | 0 | 0 | — |
case-03 | fail→fail | 6,765 | 6,135 | -9% | 1 | 1 | 0% | 283 | 508 | +80% | 0 | 0 | — |
case-04 | pass→pass | 15,781 | 15,247 | -3% | 1 | 1 | 0% | 2,588 | 2,786 | +8% | 0 | 0 | — |
case-05 | pass→pass | 14,872 | 20,748 | +40% | 1 | 1 | 0% | 3,132 | 4,937 | +58% | 0 | 0 | — |
case-06 | pass→pass | 7,853 | 6,089 | -22% | 1 | 1 | 0% | 1,145 | 1,232 | +8% | 0 | 0 | — |
case-08 | fail→pass | 12,104 | 2,079 | -83% | 1 | 1 | 0% | 2,014 | 616 | -69% | 0 | 0 | — |
case-09 | fail→pass | 10,818 | 2,164 | -80% | 1 | 1 | 0% | 1,728 | 573 | -67% | 0 | 0 | — |
case-10 | pass→pass | 9,678 | 3,851 | -60% | 1 | 1 | 0% | 1,455 | 885 | -39% | 0 | 0 | — |
case-11 | fail→pass | 12,671 | 1,789 | -86% | 1 | 1 | 0% | 2,005 | 508 | -75% | 0 | 0 | — |
case-22 | fail→pass | 18,188 | 3,456 | -81% | 1 | 1 | 0% | 3,295 | 668 | -80% | 0 | 0 | — |
case-12 | fail→pass | 10,819 | 1,749 | -84% | 1 | 1 | 0% | 1,734 | 473 | -73% | 0 | 0 | — |
case-13 | pass→pass | 13,882 | 5,223 | -62% | 1 | 1 | 0% | 2,244 | 1,061 | -53% | 0 | 0 | — |
case-14 | fail→pass | 10,773 | 1,970 | -82% | 1 | 1 | 0% | 1,635 | 459 | -72% | 0 | 0 | — |
case-15 | fail→pass | 16,057 | 1,489 | -91% | 1 | 1 | 0% | 2,563 | 460 | -82% | 0 | 0 | — |
case-17 | fail→pass | 8,029 | 2,339 | -71% | 1 | 1 | 0% | 1,072 | 582 | -46% | 0 | 0 | — |
case-18 | fail→pass | 11,688 | 2,773 | -76% | 1 | 1 | 0% | 1,638 | 550 | -66% | 0 | 0 | — |
case-19 | pass→pass | 12,712 | 2,023 | -84% | 1 | 1 | 0% | 1,894 | 537 | -72% | 0 | 0 | — |
case-20 | pass→pass | 9,544 | 1,328 | -86% | 1 | 1 | 0% | 1,604 | 426 | -73% | 0 | 0 | — |
case-21 | fail→pass | 17,816 | 10,117 | -43% | 1 | 1 | 0% | 1,281 | 990 | -23% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 19 counted toward the lift figure. The other 3 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +55 percentage points is the difference between those two pass rates over the 19 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/13/2026 | +41% |
Other measured skills in the registry, with their headline benchmark lift.