Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Define gate criteria for each stage, evaluate candidates at each gate, and render go/kill/recycle decisions with evidence.
.claude/skills/yogsoth-ai-staged-gate-evaluation/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-19 | ✗→✓ | ▲ Improved | 33% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -5% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 85% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 16% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 99% | 0% |
Apply Stage-Gate methodology to evaluate a candidate at defined decision points. Each gate has explicit criteria, and the candidate receives a GO (proceed), KILL (terminate), or RECYCLE (rework and re-evaluate) verdict based on evidence.
gate-criteria-definition SOP for each stage gate.gate-judgment SOP at each gate.feasibility-synthesis SOP.| SOP | Stage | Purpose | |-----|-------|---------| | gate-criteria-definition | 1 | Define criteria and pass thresholds | | gate-judgment | 2 | Evaluate and render verdict | | feasibility-synthesis | 3 | Synthesize into final recommendation |
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | feasibility-synthesis | Synthesize all assessments into a feasibility matrix, recommendation, and risk summary. | | gate-criteria-definition | Define gate criteria and pass thresholds for a specific stage in the Stage-Gate process. | | gate-judgment | Evaluate a candidate against gate criteria and render GO/KILL/RECYCLE verdict with evidence. |
<!-- END available-tables (generated) -->
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-19 | fail→pass | 23,485 | 26,053 | +11% | 1 | 1 | 0% | 3,073 | 4,097 | +33% | 0 | 0 | — |
case-01 | fail→pass | 45,888 | 40,680 | -11% | 1 | 1 | 0% | 6,553 | 6,236 | -5% | 0 | 0 | — |
case-02 | pass→pass | 18,324 | 38,660 | +111% | 1 | 1 | 0% | 2,794 | 5,323 | +91% | 0 | 0 | — |
case-03 | fail→fail | 37,215 | 33,771 | -9% | 1 | 1 | 0% | 5,239 | 5,478 | +5% | 0 | 0 | — |
case-04 | pass→pass | 34,996 | 32,017 | -9% | 1 | 1 | 0% | 5,479 | 4,999 | -9% | 0 | 0 | — |
case-05 | fail→fail | 25,014 | 46,526 | +86% | 1 | 1 | 0% | 2,946 | 7,493 | +154% | 0 | 0 | — |
case-06 | fail→pass | 52,039 | 40,563 | -22% | 1 | 1 | 0% | 3,503 | 6,487 | +85% | 0 | 0 | — |
case-07 | fail→pass | 27,873 | 27,477 | -1% | 1 | 1 | 0% | 3,494 | 4,048 | +16% | 0 | 0 | — |
case-08 | fail→fail | 23,443 | 30,511 | +30% | 1 | 1 | 0% | 2,844 | 4,911 | +73% | 0 | 0 | — |
case-09 | fail→fail | 31,882 | 47,626 | +49% | 1 | 1 | 0% | 5,034 | 7,064 | +40% | 0 | 0 | — |
case-10 | fail→fail | 24,781 | 20,526 | -17% | 1 | 1 | 0% | 3,198 | 3,885 | +21% | 0 | 0 | — |
case-11 | fail→pass | 23,327 | 32,079 | +38% | 1 | 1 | 0% | 2,752 | 5,477 | +99% | 0 | 0 | — |
case-12 | fail→pass | 44,131 | 50,135 | +14% | 1 | 1 | 0% | 6,581 | 7,963 | +21% | 0 | 0 | — |
case-13 | fail→pass | 21,285 | 33,864 | +59% | 1 | 1 | 0% | 2,535 | 5,176 | +104% | 0 | 0 | — |
case-14 | fail→pass | 51,660 | 37,749 | -27% | 1 | 1 | 0% | 1,423 | 5,609 | +294% | 0 | 0 | — |
case-15 | fail→fail | 25,715 | 28,783 | +12% | 1 | 1 | 0% | 3,331 | 4,521 | +36% | 0 | 0 | — |
case-16 | fail→pass | 16,163 | 21,424 | +33% | 1 | 1 | 0% | 2,523 | 3,237 | +28% | 0 | 0 | — |
case-17 | fail→pass | 22,100 | 23,908 | +8% | 1 | 1 | 0% | 2,773 | 3,628 | +31% | 0 | 0 | — |
case-18 | fail→fail | 18,715 | 26,263 | +40% | 1 | 1 | 0% | 2,979 | 3,871 | +30% | 0 | 0 | — |
case-20 | pass→fail | 29,880 | 37,070 | +24% | 1 | 1 | 0% | 4,457 | 7,007 | +57% | 0 | 0 | — |
case-21 | pass→fail | 18,953 | 35,877 | +89% | 1 | 1 | 0% | 3,634 | 6,451 | +78% | 0 | 0 | — |
case-22 | pass→fail | 21,033 | 45,937 | +118% | 1 | 1 | 0% | 3,172 | 8,743 | +176% | 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 21 counted toward the lift figure. The other 1 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 +32 percentage points is the difference between those two pass rates over the 21 comparable cases. 3 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.