Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use during planning, implementation, PR review, or /qv-qip-triage when a change may affect public SDK API, native dependency, plugin contract, model registry contract, runtime, transport, storage, release flow, deployment, security, NFR, or technical principles.
.claude/skills/tetherto-qv-qip-triage/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | 61% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -51% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -18% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -11% | 0% |
| case-20 | ✗→✓ | ▲ Improved | -16% | 0% |
Conservatively decide whether a change needs a QIP before deeper implementation or merge recommendation.
Use when:
/qv-qip-triageDo NOT use for:
qv-qip-create)qv-qip-review)qv-qip-createWhen no trigger fires:
textNo architectural significance trigger clearly applies. Proceed with normal team review.
When a trigger fires:
textTrigger: <trigger name> Why: <one or two sentences> QIP-worthy points: - <exact point that needs proposal review> Not QIP-worthy on its own: - <implementation detail or ordinary work> Proposal count: <One QIP: <scope name> / Multiple QIPs: <scope names and why split>> This looks technically significant because it changes <trigger>. I recommend drafting the proposal(s) above before going deeper, so the affected people can review the direction early. Want me to start a QIP draft from what we know?
gh pr diff| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-05 | pass→pass | 14,851 | 8,496 | -43% | 1 | 1 | 0% | 2,461 | 2,229 | -9% | 0 | 0 | — |
case-01 | fail→fail | 11,865 | 4,236 | -64% | 1 | 1 | 0% | 1,977 | 877 | -56% | 0 | 0 | — |
case-02 | fail→fail | 9,447 | 4,925 | -48% | 1 | 1 | 0% | 1,533 | 942 | -39% | 0 | 0 | — |
case-03 | pass→pass | 8,616 | 8,717 | +1% | 1 | 1 | 0% | 1,507 | 2,204 | +46% | 0 | 0 | — |
case-04 | pass→fail | 4,523 | 3,816 | -16% | 1 | 1 | 0% | 742 | 856 | +15% | 0 | 0 | — |
case-06 | fail→pass | 7,965 | 8,275 | +4% | 1 | 1 | 0% | 1,300 | 2,097 | +61% | 0 | 0 | — |
case-07 | pass→fail | 8,660 | 4,230 | -51% | 1 | 1 | 0% | 1,255 | 843 | -33% | 0 | 0 | — |
case-08 | fail→pass | 19,492 | 32,569 | +67% | 1 | 1 | 0% | 3,233 | 1,576 | -51% | 0 | 0 | — |
case-09 | pass→pass | 17,110 | 7,008 | -59% | 1 | 1 | 0% | 2,596 | 1,824 | -30% | 0 | 0 | — |
case-10 | fail→pass | 16,723 | 10,554 | -37% | 1 | 1 | 0% | 2,535 | 2,088 | -18% | 0 | 0 | — |
case-11 | fail→fail | 6,205 | 4,908 | -21% | 1 | 1 | 0% | 249 | 905 | +263% | 0 | 0 | — |
case-12 | fail→fail | 27,228 | 4,316 | -84% | 1 | 1 | 0% | 2,307 | 843 | -63% | 0 | 0 | — |
case-13 | fail→fail | 8,918 | 4,012 | -55% | 1 | 1 | 0% | 1,473 | 870 | -41% | 0 | 0 | — |
case-14 | pass→pass | 15,747 | 10,018 | -36% | 1 | 1 | 0% | 2,458 | 1,926 | -22% | 0 | 0 | — |
case-15 | fail→fail | 12,716 | 4,037 | -68% | 1 | 1 | 0% | 2,001 | 834 | -58% | 0 | 0 | — |
case-21 | pass→fail | 11,489 | 4,566 | -60% | 1 | 1 | 0% | 1,751 | 792 | -55% | 0 | 0 | — |
case-16 | fail→fail | 12,854 | 2,522 | -80% | 1 | 1 | 0% | 2,020 | 1,005 | -50% | 0 | 0 | — |
case-17 | fail→pass | 14,887 | 7,744 | -48% | 1 | 1 | 0% | 2,244 | 1,988 | -11% | 0 | 0 | — |
case-18 | pass→pass | 13,347 | 5,983 | -55% | 1 | 1 | 0% | 2,166 | 1,622 | -25% | 0 | 0 | — |
case-19 | fail→fail | 16,349 | 7,354 | -55% | 1 | 1 | 0% | 2,343 | 944 | -60% | 0 | 0 | — |
case-20 | fail→pass | 19,331 | 11,029 | -43% | 1 | 1 | 0% | 2,903 | 2,431 | -16% | 0 | 0 | — |
case-22 | fail→pass | 20,824 | 9,478 | -54% | 1 | 1 | 0% | 946 | 1,925 | +103% | 0 | 0 | — |
case-23 | fail→pass | 19,228 | 7,605 | -60% | 1 | 1 | 0% | 3,049 | 1,907 | -37% | 0 | 0 | — |
case-24 | fail→fail | 12,364 | 8,609 | -30% | 1 | 1 | 0% | 1,878 | 1,219 | -35% | 0 | 0 | — |
case-25 | fail→fail | 4,839 | 7,409 | +53% | 1 | 1 | 0% | 804 | 1,143 | +42% | 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. 25 cases were attempted, and 12 counted toward the lift figure. The other 13 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 +16 percentage points is the difference between those two pass rates over the 12 comparable cases. 6 cases got worse with the skill loaded, and they are included in that figure.
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.
Other measured skills in the registry, with their headline benchmark lift.