Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Handles Git operations with human-style commits (no AI markers). Use when user mentions git, commits, committing code, pushing changes, or wants natural developer-style commit messages. Never includes AI attribution or automated markers.
.claude/skills/dicklesworthstone-github/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 427% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 130% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 27% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 43% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 89% | 0% |
You are a specialized agent for handling Git operations with a focus on creating natural, human-written commits.
CRITICAL: All commits must look like they were written by a human developer. Absolutely NO AI markers, no overly formal language, no automated-sounding messages.
For bigger changes, use this format:
Short summary (50 chars or less)
Longer explanation if needed. Keep it casual and to the point.
- Can use bullets for multiple changes
- Don't be too formal
- Sound like you're explaining to a teammateWhen committing:
--no-verify unless explicitly askedBug fix:
New feature:
Refactoring:
Dependencies:
Documentation:
Work in progress:
❌ Never include:
When the user asks for git operations, handle everything smoothly and make commits that blend in with their repository's history.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 2,818 | 29,031 | +930% | 1 | 1 | 0% | 359 | 1,892 | +427% | 0 | 0 | — |
case-02 | fail→fail | 4,597 | 4,515 | -2% | 1 | 1 | 0% | 319 | 1,902 | +496% | 0 | 0 | — |
case-19 | pass→pass | 5,868 | 3,605 | -39% | 1 | 1 | 0% | 1,010 | 1,840 | +82% | 0 | 0 | — |
case-03 | pass→fail | 7,130 | 7,841 | +10% | 1 | 1 | 0% | 1,069 | 2,407 | +125% | 0 | 0 | — |
case-04 | pass→pass | 11,530 | 9,585 | -17% | 1 | 1 | 0% | 2,035 | 2,644 | +30% | 0 | 0 | — |
case-05 | pass→pass | 11,356 | 9,376 | -17% | 1 | 1 | 0% | 1,916 | 2,317 | +21% | 0 | 0 | — |
case-06 | fail→pass | 5,336 | 6,180 | +16% | 1 | 1 | 0% | 995 | 2,292 | +130% | 0 | 0 | — |
case-07 | fail→pass | 7,768 | 4,376 | -44% | 1 | 1 | 0% | 1,481 | 1,887 | +27% | 0 | 0 | — |
case-08 | fail→pass | 8,345 | 6,373 | -24% | 1 | 1 | 0% | 1,482 | 2,119 | +43% | 0 | 0 | — |
case-09 | pass→pass | 6,143 | 3,700 | -40% | 1 | 1 | 0% | 966 | 1,842 | +91% | 0 | 0 | — |
case-10 | fail→pass | 5,438 | 3,732 | -31% | 1 | 1 | 0% | 932 | 1,760 | +89% | 0 | 0 | — |
case-11 | fail→pass | 5,036 | 3,698 | -27% | 1 | 1 | 0% | 875 | 1,811 | +107% | 0 | 0 | — |
case-12 | pass→pass | 7,362 | 3,614 | -51% | 1 | 1 | 0% | 1,066 | 1,705 | +60% | 0 | 0 | — |
case-13 | pass→pass | 9,391 | 5,214 | -44% | 1 | 1 | 0% | 1,497 | 2,160 | +44% | 0 | 0 | — |
case-14 | pass→pass | 7,921 | 5,424 | -32% | 1 | 1 | 0% | 1,257 | 2,054 | +63% | 0 | 0 | — |
case-20 | pass→pass | 9,891 | 6,622 | -33% | 1 | 1 | 0% | 1,551 | 2,180 | +41% | 0 | 0 | — |
case-15 | fail→fail | 9,593 | 6,101 | -36% | 1 | 1 | 0% | 1,750 | 2,141 | +22% | 0 | 0 | — |
case-16 | pass→pass | 6,463 | 4,459 | -31% | 1 | 1 | 0% | 856 | 1,819 | +113% | 0 | 0 | — |
case-17 | pass→pass | 6,272 | 3,650 | -42% | 1 | 1 | 0% | 787 | 1,696 | +116% | 0 | 0 | — |
case-18 | pass→pass | 7,032 | 3,943 | -44% | 1 | 1 | 0% | 971 | 1,828 | +88% | 0 | 0 | — |
case-21 | pass→pass | 7,956 | 3,901 | -51% | 1 | 1 | 0% | 1,214 | 1,789 | +47% | 0 | 0 | — |
case-22 | fail→fail | 14,520 | 11,783 | -19% | 1 | 1 | 0% | 2,575 | 3,252 | +26% | 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 21 counted toward the lift figure. The other 1 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 +23 percentage points is the difference between those two pass rates over the 21 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.