Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Audit herdr release readiness by comparing commits since the base release against next-release changelog and docs. Use when asked to run or apply the repo's pre-release audit, validate docs/next before release, inspect issue refs that release CI will close, or finalize release docs for herdr.
.claude/skills/herdrdev-herdr-pre-release-audit/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -72% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -49% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -76% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -72% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -68% | 0% |
Use this skill only inside the herdr repository.
Read references/pre-release-audit.md and follow its workflow. Treat it as the source of truth for:
docs/next/CHANGELOG.mddocs/next/README.md and staged website docsskills/herdr/SKILL.md against shipped CLI and agent-control behaviorjust pre-release-check or its component checksjust bench-render-scaleDo not edit files during the audit unless the user explicitly asks to apply fixes. When applying fixes, keep changes scoped to the files named in the reference workflow.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 3,077 | 5,049 | +64% | 1 | 1 | 0% | 278 | 416 | +50% | 0 | 0 | — |
case-02 | fail→fail | 4,760 | 5,124 | +8% | 1 | 1 | 0% | 223 | 325 | +46% | 0 | 0 | — |
case-03 | fail→fail | 6,534 | 3,913 | -40% | 1 | 1 | 0% | 282 | 337 | +20% | 0 | 0 | — |
case-04 | fail→fail | 4,979 | 3,187 | -36% | 1 | 1 | 0% | 736 | 446 | -39% | 0 | 0 | — |
case-05 | fail→fail | 5,148 | 6,272 | +22% | 1 | 1 | 0% | 714 | 527 | -26% | 0 | 0 | — |
case-06 | fail→fail | 7,228 | 4,851 | -33% | 1 | 1 | 0% | 1,237 | 465 | -62% | 0 | 0 | — |
case-07 | fail→pass | 10,458 | 2,092 | -80% | 1 | 1 | 0% | 1,820 | 507 | -72% | 0 | 0 | — |
case-08 | fail→pass | 8,784 | 3,418 | -61% | 1 | 1 | 0% | 1,433 | 729 | -49% | 0 | 0 | — |
case-09 | pass→pass | 6,271 | 3,034 | -52% | 1 | 1 | 0% | 989 | 684 | -31% | 0 | 0 | — |
case-10 | fail→pass | 25,679 | 1,916 | -93% | 1 | 1 | 0% | 2,174 | 511 | -76% | 0 | 0 | — |
case-11 | fail→pass | 12,831 | 2,084 | -84% | 1 | 1 | 0% | 1,487 | 420 | -72% | 0 | 0 | — |
case-12 | fail→pass | 8,549 | 1,453 | -83% | 1 | 1 | 0% | 1,390 | 447 | -68% | 0 | 0 | — |
case-13 | fail→pass | 10,653 | 3,897 | -63% | 1 | 1 | 0% | 1,599 | 669 | -58% | 0 | 0 | — |
case-14 | fail→pass | 10,271 | 3,744 | -64% | 1 | 1 | 0% | 1,327 | 811 | -39% | 0 | 0 | — |
case-15 | pass→pass | 14,946 | 7,358 | -51% | 1 | 1 | 0% | 2,389 | 953 | -60% | 0 | 0 | — |
case-16 | pass→pass | 7,095 | 2,414 | -66% | 1 | 1 | 0% | 1,136 | 484 | -57% | 0 | 0 | — |
case-17 | pass→pass | 7,439 | 3,049 | -59% | 1 | 1 | 0% | 1,065 | 611 | -43% | 0 | 0 | — |
case-18 | pass→pass | 8,633 | 9,214 | +7% | 1 | 1 | 0% | 1,467 | 1,547 | +5% | 0 | 0 | — |
case-19 | pass→pass | 6,914 | 3,176 | -54% | 1 | 1 | 0% | 933 | 652 | -30% | 0 | 0 | — |
case-20 | fail→pass | 7,869 | 3,068 | -61% | 1 | 1 | 0% | 1,165 | 526 | -55% | 0 | 0 | — |
case-21 | pass→pass | 11,396 | 2,657 | -77% | 1 | 1 | 0% | 1,681 | 551 | -67% | 0 | 0 | — |
case-22 | fail→pass | 10,151 | 3,414 | -66% | 1 | 1 | 0% | 1,551 | 573 | -63% | 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 17 counted toward the lift figure. The other 5 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 17 comparable cases. 2 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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/10/2026 | +9% |
Other measured skills in the registry, with their headline benchmark lift.