Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Automatically draft commit messages by analyzing git status and staged changes
.claude/skills/aiskillstore-commit-drafter/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -19% | 0% |
| case-16 | ✓→✗ | ▼ Worse | 4% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 91% | 0% |
| case-05 | ✓→✓ | ＝ Same ✓ | 45% | 0% |
| case-06 | ✓→✓ | ＝ Same ✓ | 2% | 0% |
Automatically generates commit messages by analyzing your staged changes.
Just say:
The skill will automatically:
Unlike template-based commit tools that just output placeholders, this skill provides real git context to Claude (the AI), who then writes a meaningful commit message based on understanding the actual code changes.
The script outputs:
Claude then reads this context and writes a commit message that actually describes what changed and why.
No more staring at git diff trying to write a commit message. No more useless templates with TODO] placeholders. Claude reads your changes, understands them, and drafts a meaningful message automatically.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 3,161 | 2,554 | -19% | 1 | 1 | 0% | 529 | 579 | +9% | 0 | 0 | — |
case-02 | fail→fail | 2,582 | 4,784 | +85% | 1 | 1 | 0% | 377 | 998 | +165% | 0 | 0 | — |
case-03 | fail→fail | 2,238 | 2,994 | +34% | 1 | 1 | 0% | 289 | 569 | +97% | 0 | 0 | — |
case-04 | pass→pass | 2,961 | 3,678 | +24% | 1 | 1 | 0% | 437 | 835 | +91% | 0 | 0 | — |
case-05 | pass→pass | 11,498 | 18,305 | +59% | 1 | 1 | 0% | 1,926 | 2,790 | +45% | 0 | 0 | — |
case-06 | pass→pass | 6,934 | 5,754 | -17% | 1 | 1 | 0% | 1,376 | 1,407 | +2% | 0 | 0 | — |
case-07 | fail→pass | 6,041 | 3,272 | -46% | 1 | 1 | 0% | 1,038 | 841 | -19% | 0 | 0 | — |
case-08 | pass→pass | 5,840 | 3,873 | -34% | 1 | 1 | 0% | 979 | 986 | +1% | 0 | 0 | — |
case-09 | pass→pass | 5,887 | 3,907 | -34% | 1 | 1 | 0% | 997 | 1,002 | +1% | 0 | 0 | — |
case-10 | pass→pass | 3,712 | 3,954 | +7% | 1 | 1 | 0% | 636 | 990 | +56% | 0 | 0 | — |
case-11 | pass→pass | 5,052 | 3,313 | -34% | 1 | 1 | 0% | 958 | 907 | -5% | 0 | 0 | — |
case-12 | pass→pass | 5,451 | 3,794 | -30% | 1 | 1 | 0% | 960 | 968 | +1% | 0 | 0 | — |
case-13 | pass→pass | 5,494 | 3,862 | -30% | 1 | 1 | 0% | 901 | 987 | +10% | 0 | 0 | — |
case-14 | pass→pass | 5,025 | 3,843 | -24% | 1 | 1 | 0% | 902 | 874 | -3% | 0 | 0 | — |
case-15 | pass→pass | 5,506 | 3,705 | -33% | 1 | 1 | 0% | 949 | 992 | +5% | 0 | 0 | — |
case-16 | pass→fail | 5,377 | 3,936 | -27% | 1 | 1 | 0% | 906 | 945 | +4% | 0 | 0 | — |
case-17 | pass→pass | 4,098 | 3,842 | -6% | 1 | 1 | 0% | 780 | 961 | +23% | 0 | 0 | — |
case-18 | pass→pass | 5,972 | 3,860 | -35% | 1 | 1 | 0% | 1,018 | 958 | -6% | 0 | 0 | — |
case-19 | pass→pass | 5,012 | 3,632 | -28% | 1 | 1 | 0% | 838 | 880 | +5% | 0 | 0 | — |
case-20 | pass→pass | 4,592 | 3,546 | -23% | 1 | 1 | 0% | 852 | 923 | +8% | 0 | 0 | — |
case-21 | pass→pass | 7,645 | 3,383 | -56% | 1 | 1 | 0% | 1,251 | 878 | -30% | 0 | 0 | — |
case-22 | pass→pass | 4,859 | 3,663 | -25% | 1 | 1 | 0% | 886 | 962 | +9% | 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 0 percentage points is the difference between those two pass rates over the 22 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.