Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Stage, review, and commit files following Penpot commit conventions.
.claude/skills/penpot-create-commit/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | -47% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 285% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -71% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-07 | ✓→✗ | ▼ Worse | 13% | 0% |
Produce a git commit that follows Penpot's commit message conventions. This skill owns the commit format, staging review, and safety checks — it does not implement features or push.
Before drafting any commit, read mem:workflow/creating-commits end-to-end. It is the authoritative source for the commit message format, the emoji menu, subject/body limits, and the AI-assisted-by trailer. Follow it exactly.
not eyeball. Exceptions: Signed-off-by: / AI-assisted-by: trailers and lines carrying a URL. This is the rule agents skip most often.
./scripts/check-commit and require exit code 0. It mechanicallychecks rules 1–3. A non-zero exit is a hard blocker: fix the message and re-commit. Never report the commit as done with a failing checker.
confirmation.
git diff --staged to review the content. If you see secrets (API keys,tokens, passwords, private keys, .env values), debug prints, or anything that does not match the stated intent, STOP and tell the user before committing.
at 76 characters per line, and run: bash git commit -m "<subject>" -m "<body>" (or git commit -F - if the body has unusual characters).
bash ./scripts/check-commit If it fails, amend the message (git commit --amend) until it passes. Do not finish with a failing checker.
AI-assisted-by trailer value is provided by the calling context — useit verbatim.
git reset, git checkout, git restore, git clean, or rm.--author. Author identity comes from the local git config.--no-verify) unless explicitly asked.scripts/check-commit verification step (Iron Rule 4).| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 7,556 | 5,681 | -25% | 1 | 1 | 0% | 1,034 | 914 | -12% | 0 | 0 | — |
case-02 | fail→fail | 6,708 | 6,399 | -5% | 1 | 1 | 0% | 832 | 946 | +14% | 0 | 0 | — |
case-03 | fail→fail | 3,591 | 7,396 | +106% | 1 | 1 | 0% | 350 | 953 | +172% | 0 | 0 | — |
case-04 | pass→pass | 4,750 | 3,768 | -21% | 1 | 1 | 0% | 670 | 1,092 | +63% | 0 | 0 | — |
case-05 | pass→pass | 7,253 | 3,800 | -48% | 1 | 1 | 0% | 1,014 | 1,168 | +15% | 0 | 0 | — |
case-06 | fail→fail | 16,542 | 7,116 | -57% | 1 | 1 | 0% | 2,625 | 956 | -64% | 0 | 0 | — |
case-07 | pass→fail | 12,760 | 7,835 | -39% | 1 | 1 | 0% | 1,002 | 1,129 | +13% | 0 | 0 | — |
case-08 | pass→pass | 20,920 | 20,121 | -4% | 1 | 1 | 0% | 3,802 | 4,561 | +20% | 0 | 0 | — |
case-09 | fail→pass | 11,179 | 2,090 | -81% | 1 | 1 | 0% | 1,665 | 879 | -47% | 0 | 0 | — |
case-10 | fail→pass | 4,824 | 7,341 | +52% | 1 | 1 | 0% | 448 | 1,726 | +285% | 0 | 0 | — |
case-11 | pass→pass | 7,419 | 3,027 | -59% | 1 | 1 | 0% | 738 | 1,011 | +37% | 0 | 0 | — |
case-12 | fail→pass | 18,662 | 3,393 | -82% | 1 | 1 | 0% | 3,251 | 937 | -71% | 0 | 0 | — |
case-13 | fail→fail | 3,862 | 8,039 | +108% | 1 | 1 | 0% | 269 | 1,037 | +286% | 0 | 0 | — |
case-14 | fail→fail | 4,966 | 9,437 | +90% | 1 | 1 | 0% | 560 | 1,092 | +95% | 0 | 0 | — |
case-15 | fail→pass | 7,123 | 2,819 | -60% | 1 | 1 | 0% | 973 | 966 | -1% | 0 | 0 | — |
case-16 | pass→pass | 8,466 | 2,890 | -66% | 1 | 1 | 0% | 1,053 | 1,056 | +0% | 0 | 0 | — |
case-17 | pass→pass | 8,279 | 3,129 | -62% | 1 | 1 | 0% | 1,180 | 1,109 | -6% | 0 | 0 | — |
case-18 | fail→fail | 6,141 | 15,206 | +148% | 1 | 1 | 0% | 700 | 1,047 | +50% | 0 | 0 | — |
case-19 | fail→fail | 9,063 | 11,803 | +30% | 1 | 1 | 0% | 466 | 1,020 | +119% | 0 | 0 | — |
case-20 | fail→fail | 7,181 | 7,669 | +7% | 1 | 1 | 0% | 963 | 1,128 | +17% | 0 | 0 | — |
case-21 | pass→pass | 6,494 | 5,614 | -14% | 1 | 1 | 0% | 974 | 1,528 | +57% | 0 | 0 | — |
case-22 | pass→pass | 15,643 | 4,024 | -74% | 1 | 1 | 0% | 1,367 | 1,135 | -17% | 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 12 counted toward the lift figure. The other 10 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 +14 percentage points is the difference between those two pass rates over the 12 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.