Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Prepare @indielayer/ui changes for release with changesets and conventional commits. Use when versioning, publishing, changelog, semver, or before merging user-facing library changes.
.claude/skills/indielayer-release/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -28% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -45% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -54% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -61% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -70% | 0% |
For user-facing packages/ui changes:
bashpnpm changeset
Choose bump type (patch/minor/major) and write a short summary for the changelog.
Follow .github/COMMIT_CONVENTION.md:
feat(scope): ... — minor bump, appears in changelogfix(scope): ... — patch bumpperf(scope): ... — patch bumpdocs, chore, test, etc. — no version bump unless changeset addedDo not bump packages/ui/package.json version in the feature PR; Changesets handles release.
bashpnpm lint:ui pnpm typecheck pnpm test:ci pnpm build
Ensure a changeset file exists under .changeset/ for library changes.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 19,763 | 6,460 | -67% | 1 | 1 | 0% | 1,959 | 1,408 | -28% | 0 | 0 | — |
case-02 | fail→fail | 14,061 | 5,297 | -62% | 1 | 1 | 0% | 2,000 | 1,098 | -45% | 0 | 0 | — |
case-03 | fail→pass | 18,803 | 6,609 | -65% | 1 | 1 | 0% | 2,569 | 1,425 | -45% | 0 | 0 | — |
case-04 | pass→pass | 10,671 | 3,057 | -71% | 1 | 1 | 0% | 1,868 | 751 | -60% | 0 | 0 | — |
case-05 | fail→pass | 9,855 | 4,633 | -53% | 1 | 1 | 0% | 1,389 | 643 | -54% | 0 | 0 | — |
case-06 | fail→pass | 7,742 | 2,269 | -71% | 1 | 1 | 0% | 1,152 | 451 | -61% | 0 | 0 | — |
case-07 | pass→pass | 5,708 | 3,219 | -44% | 1 | 1 | 0% | 799 | 623 | -22% | 0 | 0 | — |
case-08 | pass→pass | 5,171 | 3,120 | -40% | 1 | 1 | 0% | 821 | 624 | -24% | 0 | 0 | — |
case-09 | pass→pass | 5,798 | 2,694 | -54% | 1 | 1 | 0% | 908 | 534 | -41% | 0 | 0 | — |
case-10 | pass→pass | 8,346 | 4,142 | -50% | 1 | 1 | 0% | 1,398 | 860 | -38% | 0 | 0 | — |
case-11 | fail→pass | 9,530 | 2,769 | -71% | 1 | 1 | 0% | 1,336 | 407 | -70% | 0 | 0 | — |
case-12 | fail→pass | 5,889 | 1,658 | -72% | 1 | 1 | 0% | 941 | 429 | -54% | 0 | 0 | — |
case-13 | fail→fail | 7,394 | 2,436 | -67% | 1 | 1 | 0% | 1,316 | 544 | -59% | 0 | 0 | — |
case-14 | pass→pass | 6,467 | 2,232 | -65% | 1 | 1 | 0% | 1,056 | 466 | -56% | 0 | 0 | — |
case-15 | pass→pass | 6,919 | 2,397 | -65% | 1 | 1 | 0% | 1,126 | 410 | -64% | 0 | 0 | — |
case-16 | pass→pass | 6,575 | 3,042 | -54% | 1 | 1 | 0% | 1,235 | 682 | -45% | 0 | 0 | — |
case-17 | fail→fail | 6,797 | 2,084 | -69% | 1 | 1 | 0% | 1,268 | 437 | -66% | 0 | 0 | — |
case-18 | pass→pass | 14,652 | 4,312 | -71% | 1 | 1 | 0% | 1,580 | 712 | -55% | 0 | 0 | — |
case-19 | pass→pass | 4,088 | 2,774 | -32% | 1 | 1 | 0% | 657 | 461 | -30% | 0 | 0 | — |
case-20 | fail→fail | 8,407 | 5,827 | -31% | 1 | 1 | 0% | 1,570 | 963 | -39% | 0 | 0 | — |
case-21 | fail→fail | 10,987 | 9,356 | -15% | 1 | 1 | 0% | 2,014 | 2,201 | +9% | 0 | 0 | — |
case-22 | fail→fail | 12,243 | 6,516 | -47% | 1 | 1 | 0% | 1,707 | 1,245 | -27% | 0 | 0 | — |
case-23 | fail→fail | 4,757 | 3,050 | -36% | 1 | 1 | 0% | 703 | 615 | -13% | 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. 23 cases were attempted. The headline lift of +26 percentage points is the difference between those two pass rates over the 23 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.