Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Analyzes staged and unstaged changes, generates a commit message that summarizes the "why" not just the "what", and creates the commit. Follows conventional commit format when the project uses it.
.claude/skills/miosa-osa-commit/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-13 | ✗→✓ | ▲ Improved | 402% | 0% |
| case-20 | ✗→✓ | ▲ Improved | -41% | 0% |
| case-07 | ✓→✗ | ▼ Worse | -14% | 0% |
| case-01 | ✓→✗ | ▼ Worse | -30% | 0% |
| case-02 | ✓→✗ | ▼ Worse | -59% | 0% |
> Stage changes, generate a meaningful commit message, and commit.
/commit [--all] [--amend] [-m "message"]Analyzes staged and unstaged changes, generates a commit message that summarizes the "why" not just the "what", and creates the commit. Follows conventional commit format when the project uses it.
git diff --staged and git status.--all) -- add all modified tracked files.git status to confirm clean state.bash# Commit staged changes with auto-generated message /commit # Stage all changes and commit /commit --all # Commit with explicit message /commit -m "fix: resolve race condition in session handler" # Amend the last commit /commit --amend
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-17 | fail→fail | 3,697 | 2,912 | -21% | 1 | 1 | 0% | 153 | 457 | +199% | 0 | 0 | — |
case-07 | pass→fail | 5,251 | 2,977 | -43% | 1 | 1 | 0% | 672 | 575 | -14% | 0 | 0 | — |
case-01 | pass→fail | 3,833 | 4,449 | +16% | 1 | 1 | 0% | 607 | 427 | -30% | 0 | 0 | — |
case-02 | pass→fail | 6,535 | 6,220 | -5% | 1 | 1 | 0% | 1,239 | 509 | -59% | 0 | 0 | — |
case-03 | pass→pass | 6,166 | 18,982 | +208% | 1 | 1 | 0% | 791 | 585 | -26% | 0 | 0 | — |
case-04 | fail→fail | 3,064 | 2,966 | -3% | 1 | 1 | 0% | 264 | 531 | +101% | 0 | 0 | — |
case-05 | fail→fail | 3,944 | 30,768 | +680% | 1 | 1 | 0% | 159 | 519 | +226% | 0 | 0 | — |
case-06 | fail→fail | 14,070 | 4,112 | -71% | 1 | 1 | 0% | 2,189 | 516 | -76% | 0 | 0 | — |
case-08 | pass→pass | 4,111 | 2,168 | -47% | 1 | 1 | 0% | 682 | 525 | -23% | 0 | 0 | — |
case-09 | fail→fail | 2,601 | 2,462 | -5% | 1 | 1 | 0% | 414 | 493 | +19% | 0 | 0 | — |
case-10 | pass→pass | 7,128 | 7,695 | +8% | 1 | 1 | 0% | 738 | 1,017 | +38% | 0 | 0 | — |
case-11 | fail→fail | 3,306 | 6,204 | +88% | 1 | 1 | 0% | 289 | 502 | +74% | 0 | 0 | — |
case-12 | pass→fail | 5,639 | 5,885 | +4% | 1 | 1 | 0% | 663 | 455 | -31% | 0 | 0 | — |
case-13 | fail→pass | 14,638 | 13,412 | -8% | 1 | 1 | 0% | 369 | 1,853 | +402% | 0 | 0 | — |
case-14 | fail→fail | 2,678 | 2,644 | -1% | 1 | 1 | 0% | 251 | 535 | +113% | 0 | 0 | — |
case-15 | pass→fail | 7,766 | 5,866 | -24% | 1 | 1 | 0% | 1,331 | 553 | -58% | 0 | 0 | — |
case-16 | pass→fail | 10,367 | 5,100 | -51% | 1 | 1 | 0% | 648 | 652 | +1% | 0 | 0 | — |
case-18 | pass→pass | 6,332 | 5,415 | -14% | 1 | 1 | 0% | 969 | 914 | -6% | 0 | 0 | — |
case-19 | pass→pass | 4,339 | 10,284 | +137% | 1 | 1 | 0% | 683 | 1,587 | +132% | 0 | 0 | — |
case-20 | fail→pass | 6,201 | 2,502 | -60% | 1 | 1 | 0% | 1,120 | 663 | -41% | 0 | 0 | — |
case-21 | pass→pass | 9,058 | 8,733 | -4% | 1 | 1 | 0% | 1,342 | 1,625 | +21% | 0 | 0 | — |
case-22 | pass→pass | 7,581 | 4,074 | -46% | 1 | 1 | 0% | 1,332 | 825 | -38% | 0 | 0 | — |
case-23 | pass→fail | 1,823 | 5,870 | +222% | 1 | 1 | 0% | 303 | 560 | +85% | 0 | 0 | — |
case-24 | pass→pass | 5,214 | 11,855 | +127% | 1 | 1 | 0% | 613 | 1,239 | +102% | 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. 24 cases were attempted, and 16 counted toward the lift figure. The other 8 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 -50 percentage points is the difference between those two pass rates over the 16 comparable cases. 7 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.
Other measured skills in the registry, with their headline benchmark lift.