Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Define gate criteria and pass thresholds for a specific stage in the Stage-Gate process.
.claude/skills/yogsoth-ai-gate-criteria-definition/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 112% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 86% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 434% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 207% | 0% |
Define explicit, measurable criteria that a candidate must satisfy to pass through a specific stage gate. Each criterion has a pass threshold and evidence requirements.
Spawns a subagent that:
Gate criteria must be defined independently of the candidate being evaluated to avoid bias. A dedicated subagent ensures criteria are set based on the stage requirements, not tailored to make a specific candidate pass.
Output MUST include: at least 3 gate criteria, measurable pass threshold for each, and evidence requirements. Reject if criteria are vague or unmeasurable.
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | spawn-agent | Spawn a customized CC subagent with full MCP tool access. Used by SOPs that declare execution: subagent. |
<!-- 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-01 | fail→pass | 42,035 | 44,227 | +5% | 1 | 1 | 0% | 1,725 | 3,660 | +112% | 0 | 0 | — |
case-02 | pass→fail | 16,214 | 11,049 | -32% | 1 | 1 | 0% | 2,444 | 2,225 | -9% | 0 | 0 | — |
case-03 | fail→pass | 11,071 | 17,127 | +55% | 1 | 1 | 0% | 1,704 | 3,171 | +86% | 0 | 0 | — |
case-04 | pass→pass | 19,116 | 21,617 | +13% | 1 | 1 | 0% | 3,377 | 4,055 | +20% | 0 | 0 | — |
case-05 | fail→pass | 3,348 | 14,090 | +321% | 1 | 1 | 0% | 497 | 2,654 | +434% | 0 | 0 | — |
case-06 | pass→pass | 16,752 | 16,325 | -3% | 1 | 1 | 0% | 2,775 | 2,950 | +6% | 0 | 0 | — |
case-07 | fail→pass | 18,613 | 16,877 | -9% | 1 | 1 | 0% | 2,982 | 2,957 | -1% | 0 | 0 | — |
case-08 | pass→pass | 16,329 | 18,120 | +11% | 1 | 1 | 0% | 2,476 | 3,387 | +37% | 0 | 0 | — |
case-09 | fail→fail | 18,896 | 11,151 | -41% | 1 | 1 | 0% | 2,975 | 2,063 | -31% | 0 | 0 | — |
case-10 | fail→pass | 31,657 | 20,431 | -35% | 1 | 1 | 0% | 1,214 | 3,723 | +207% | 0 | 0 | — |
case-11 | fail→pass | 13,595 | 16,620 | +22% | 1 | 1 | 0% | 2,256 | 3,313 | +47% | 0 | 0 | — |
case-12 | fail→pass | 13,513 | 11,898 | -12% | 1 | 1 | 0% | 2,192 | 2,190 | -0% | 0 | 0 | — |
case-13 | pass→pass | 15,950 | 17,923 | +12% | 1 | 1 | 0% | 2,615 | 3,094 | +18% | 0 | 0 | — |
case-14 | pass→pass | 14,994 | 22,041 | +47% | 1 | 1 | 0% | 2,324 | 3,755 | +62% | 0 | 0 | — |
case-15 | fail→pass | 12,642 | 12,110 | -4% | 1 | 1 | 0% | 1,970 | 2,474 | +26% | 0 | 0 | — |
case-16 | fail→fail | 18,729 | 14,043 | -25% | 1 | 1 | 0% | 2,857 | 2,599 | -9% | 0 | 0 | — |
case-17 | fail→fail | 18,674 | 20,034 | +7% | 1 | 1 | 0% | 3,048 | 3,541 | +16% | 0 | 0 | — |
case-18 | fail→pass | 11,395 | 6,830 | -40% | 1 | 1 | 0% | 1,632 | 1,207 | -26% | 0 | 0 | — |
case-19 | fail→pass | 16,936 | 35,960 | +112% | 1 | 1 | 0% | 2,747 | 2,925 | +6% | 0 | 0 | — |
case-20 | pass→pass | 3,438 | 12,546 | +265% | 1 | 1 | 0% | 584 | 2,712 | +364% | 0 | 0 | — |
case-21 | pass→pass | 9,878 | 22,278 | +126% | 1 | 1 | 0% | 1,630 | 4,261 | +161% | 0 | 0 | — |
case-22 | pass→pass | 7,991 | 17,267 | +116% | 1 | 1 | 0% | 1,474 | 3,217 | +118% | 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 20 counted toward the lift figure. The other 2 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 +41 percentage points is the difference between those two pass rates over the 20 comparable cases. 1 case got worse with the skill loaded, and it is 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.