Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Проверь release readiness Gemini adapter и public CI policy. EN: release review.
.claude/skills/bilal140202-release-review/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 10% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -49% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -54% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -40% | 0% |
Validate that current release surfaces, CI, tags, metadata, and archives are coherent.
Release review validates Gemini-native config and public adapter policy.
Use before tagging, publishing, or integrating this adapter into the root tuple.
Version files, changelog, workflows, validators, tests, and release artifacts.
Command results, tag/release URL, commit SHA, and artifact validation output.
Do not certify release with NOT_PROVEN runtime, GitHub, or archive evidence.
All local validators, pytest, and release/archive hygiene checks pass.
State exact non-green gates and whether they are blockers or NOT_PROVEN.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 17,357 | 12,694 | -27% | 1 | 1 | 0% | 2,673 | 2,351 | -12% | 0 | 0 | — |
case-12 | pass→pass | 9,005 | 4,538 | -50% | 1 | 1 | 0% | 1,487 | 1,073 | -28% | 0 | 0 | — |
case-02 | fail→fail | 15,793 | 22,446 | +42% | 1 | 1 | 0% | 2,468 | 3,465 | +40% | 0 | 0 | — |
case-03 | fail→fail | 7,719 | 4,803 | -38% | 1 | 1 | 0% | 1,233 | 434 | -65% | 0 | 0 | — |
case-04 | pass→pass | 4,983 | 3,833 | -23% | 1 | 1 | 0% | 834 | 808 | -3% | 0 | 0 | — |
case-05 | pass→pass | 11,173 | 9,722 | -13% | 1 | 1 | 0% | 2,010 | 1,769 | -12% | 0 | 0 | — |
case-06 | pass→pass | 7,802 | 4,970 | -36% | 1 | 1 | 0% | 1,429 | 1,160 | -19% | 0 | 0 | — |
case-07 | fail→pass | 18,574 | 10,710 | -42% | 1 | 1 | 0% | 737 | 810 | +10% | 0 | 0 | — |
case-08 | pass→pass | 8,754 | 4,538 | -48% | 1 | 1 | 0% | 1,467 | 999 | -32% | 0 | 0 | — |
case-09 | fail→fail | 29,120 | 5,862 | -80% | 1 | 1 | 0% | 1,051 | 1,154 | +10% | 0 | 0 | — |
case-10 | fail→pass | 12,050 | 4,304 | -64% | 1 | 1 | 0% | 1,744 | 893 | -49% | 0 | 0 | — |
case-11 | fail→pass | 13,949 | 4,904 | -65% | 1 | 1 | 0% | 2,232 | 1,033 | -54% | 0 | 0 | — |
case-13 | pass→pass | 10,607 | 4,218 | -60% | 1 | 1 | 0% | 1,637 | 823 | -50% | 0 | 0 | — |
case-14 | pass→pass | 4,699 | 3,560 | -24% | 1 | 1 | 0% | 769 | 851 | +11% | 0 | 0 | — |
case-15 | fail→pass | 8,781 | 4,032 | -54% | 1 | 1 | 0% | 1,392 | 842 | -40% | 0 | 0 | — |
case-16 | fail→pass | 12,387 | 4,040 | -67% | 1 | 1 | 0% | 1,913 | 847 | -56% | 0 | 0 | — |
case-17 | fail→pass | 16,518 | 3,303 | -80% | 1 | 1 | 0% | 950 | 785 | -17% | 0 | 0 | — |
case-18 | fail→pass | 16,532 | 4,516 | -73% | 1 | 1 | 0% | 2,644 | 1,035 | -61% | 0 | 0 | — |
case-19 | pass→pass | 10,483 | 3,460 | -67% | 1 | 1 | 0% | 1,829 | 670 | -63% | 0 | 0 | — |
case-20 | pass→pass | 12,950 | 3,791 | -71% | 1 | 1 | 0% | 1,976 | 760 | -62% | 0 | 0 | — |
case-21 | fail→pass | 11,159 | 2,508 | -78% | 1 | 1 | 0% | 1,782 | 582 | -67% | 0 | 0 | — |
case-22 | fail→pass | 11,903 | 5,064 | -57% | 1 | 1 | 0% | 2,154 | 1,076 | -50% | 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 +45 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.
Other measured skills in the registry, with their headline benchmark lift.