Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Enforce conventional commit messages with commitlint. Use when a user asks to standardize commit messages, enforce commit conventions, set up commit linting in CI, or generate changelogs from commits.
.claude/skills/terminalskills-commitlint/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -17% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 25% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 9% | 0% |
| case-03 | ✓→✓ | = Same ✓ | -37% | 0% |
| case-04 | ✓→✓ | = Same ✓ | 11% | 0% |
commitlint checks commit messages against conventional commit format (type(scope): description). Pairs with husky for Git hooks and standard-version/changesets for automated changelogs.
bashnpm install -D @commitlint/cli @commitlint/config-conventional husky npx husky init echo 'npx --no -- commitlint --edit "$1"' > .husky/commit-msg
javascript// commitlint.config.js — Commit message rules export default { extends: ['@commitlint/config-conventional'], rules: { 'type-enum': [2, 'always', ['feat', 'fix', 'docs', 'style', 'refactor', 'perf', 'test', 'build', 'ci', 'chore']], 'scope-case': [2, 'always', 'kebab-case'], 'subject-max-length': [2, 'always', 72], }, }
bashgit commit -m "feat(auth): add Google OAuth login" # valid git commit -m "fix(api): handle null response from /users" # valid git commit -m "updated stuff" # rejected
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 6,999 | 4,495 | -36% | 1 | 1 | 0% | 1,646 | 1,371 | -17% | 0 | 0 | — |
case-02 | fail→pass | 10,180 | 10,141 | -0% | 1 | 1 | 0% | 1,888 | 2,355 | +25% | 0 | 0 | — |
case-03 | pass→pass | 7,581 | 2,725 | -64% | 1 | 1 | 0% | 1,308 | 822 | -37% | 0 | 0 | — |
case-04 | pass→pass | 3,365 | 1,910 | -43% | 1 | 1 | 0% | 576 | 642 | +11% | 0 | 0 | — |
case-05 | pass→pass | 7,256 | 2,812 | -61% | 1 | 1 | 0% | 1,256 | 838 | -33% | 0 | 0 | — |
case-06 | pass→pass | 7,525 | 4,515 | -40% | 1 | 1 | 0% | 1,512 | 1,256 | -17% | 0 | 0 | — |
case-07 | pass→pass | 2,071 | 2,170 | +5% | 1 | 1 | 0% | 386 | 617 | +60% | 0 | 0 | — |
case-12 | pass→pass | 3,604 | 2,052 | -43% | 1 | 1 | 0% | 764 | 662 | -13% | 0 | 0 | — |
case-08 | fail→fail | 6,970 | 5,201 | -25% | 1 | 1 | 0% | 1,356 | 1,426 | +5% | 0 | 0 | — |
case-09 | pass→pass | 6,591 | 2,390 | -64% | 1 | 1 | 0% | 1,169 | 753 | -36% | 0 | 0 | — |
case-10 | pass→pass | 6,446 | 5,362 | -17% | 1 | 1 | 0% | 1,354 | 1,359 | +0% | 0 | 0 | — |
case-11 | pass→pass | 4,342 | 3,002 | -31% | 1 | 1 | 0% | 621 | 707 | +14% | 0 | 0 | — |
case-13 | pass→pass | 6,808 | 5,292 | -22% | 1 | 1 | 0% | 1,181 | 1,162 | -2% | 0 | 0 | — |
case-14 | pass→pass | 3,598 | 2,766 | -23% | 1 | 1 | 0% | 565 | 837 | +48% | 0 | 0 | — |
case-15 | pass→pass | 2,188 | 2,335 | +7% | 1 | 1 | 0% | 373 | 759 | +103% | 0 | 0 | — |
case-16 | pass→pass | 3,667 | 3,897 | +6% | 1 | 1 | 0% | 604 | 918 | +52% | 0 | 0 | — |
case-17 | pass→pass | 11,180 | 9,371 | -16% | 1 | 1 | 0% | 1,744 | 1,788 | +3% | 0 | 0 | — |
case-18 | fail→pass | 13,057 | 12,099 | -7% | 1 | 1 | 0% | 2,302 | 2,505 | +9% | 0 | 0 | — |
case-19 | pass→pass | 10,687 | 6,225 | -42% | 1 | 1 | 0% | 2,082 | 1,556 | -25% | 0 | 0 | — |
case-20 | pass→pass | 8,737 | 6,312 | -28% | 1 | 1 | 0% | 1,594 | 1,497 | -6% | 0 | 0 | — |
case-21 | pass→pass | 9,610 | 3,308 | -66% | 1 | 1 | 0% | 1,278 | 885 | -31% | 0 | 0 | — |
case-22 | pass→pass | 7,229 | 4,528 | -37% | 1 | 1 | 0% | 1,196 | 1,132 | -5% | 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. The headline lift of +14 percentage points is the difference between those two pass rates over the 22 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.