Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Prepare PRs for review by cleaning noisy history, improving PR descriptions, and adding reviewer guidance without changing code behavior. Use for \"make this easy to review\", \"tidy this PR\", \"clean up commits\", or \"annotate the diff\".
.claude/skills/kunanonj-cursor-plugin-teamkit-make-pr-easy-to-review/SKILL.md| Model | Eval pass | Runs |
|---|---|---|
| gemini-3.6-flash | 100% | 14 |
| gemini-3.1-pro-preview | 100% | 2 |
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-13 | ✗→✓ | ▲ Improved | -15% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 45% | 0% |
| case-22 | ✓→✓ | = Same ✓ | -11% | 0% |
| case-04 | ✓→✓ | = Same ✓ | 5% | 0% |
| case-07 | ✓→✓ | = Same ✓ | 18% | 0% |
Prepare a PR so a reviewer can quickly understand the intent, important files, and risk. The default goal is reviewability without behavior changes.
Only rewrite history when the user asks for it or agrees to the plan. Before rewriting:
bashgh pr view <PR> --json title,headRefName,baseRefName,state,commits git fetch origin <headRefName> <baseRefName> ORIGINAL_TREE=$(git rev-parse origin/<headRefName>^{tree})
Good commit groupings usually follow dependency order:
After rewriting, verify content identity:
bashecho "Original tree: $ORIGINAL_TREE" echo "Current tree: $(git rev-parse HEAD^{tree})" git diff origin/<headRefName> --stat
Do not push if the tree changed unintentionally.
When code behavior should stay untouched, prefer PR description and review notes:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-02 | fail→fail | 13,177 | 5,068 | -62% | 1 | 1 | 0% | 2,273 | 794 | -65% | 0 | 0 | — |
case-22 | pass→pass | 15,799 | 12,490 | -21% | 1 | 1 | 0% | 3,112 | 2,774 | -11% | 0 | 0 | — |
case-01 | fail→fail | 13,654 | 14,044 | +3% | 1 | 1 | 0% | 2,505 | 2,973 | +19% | 0 | 0 | — |
case-03 | fail→fail | 11,384 | 8,105 | -29% | 1 | 1 | 0% | 2,176 | 2,027 | -7% | 0 | 0 | — |
case-04 | pass→pass | 11,352 | 8,249 | -27% | 1 | 1 | 0% | 1,877 | 1,979 | +5% | 0 | 0 | — |
case-05 | fail→fail | 9,556 | 6,245 | -35% | 1 | 1 | 0% | 1,708 | 1,641 | -4% | 0 | 0 | — |
case-06 | fail→fail | 11,643 | 12,165 | +4% | 1 | 1 | 0% | 1,954 | 2,401 | +23% | 0 | 0 | — |
case-07 | pass→pass | 12,298 | 11,746 | -4% | 1 | 1 | 0% | 2,046 | 2,408 | +18% | 0 | 0 | — |
case-08 | pass→pass | 9,649 | 9,446 | -2% | 1 | 1 | 0% | 1,564 | 2,064 | +32% | 0 | 0 | — |
case-09 | pass→pass | 10,132 | 4,485 | -56% | 1 | 1 | 0% | 1,668 | 1,195 | -28% | 0 | 0 | — |
case-10 | pass→pass | 8,597 | 3,484 | -59% | 1 | 1 | 0% | 1,500 | 1,112 | -26% | 0 | 0 | — |
case-11 | pass→pass | 14,298 | 10,059 | -30% | 1 | 1 | 0% | 2,356 | 2,151 | -9% | 0 | 0 | — |
case-12 | pass→pass | 10,683 | 8,038 | -25% | 1 | 1 | 0% | 1,626 | 1,709 | +5% | 0 | 0 | — |
case-13 | fail→pass | 9,166 | 4,768 | -48% | 1 | 1 | 0% | 1,579 | 1,337 | -15% | 0 | 0 | — |
case-14 | pass→pass | 11,210 | 2,374 | -79% | 1 | 1 | 0% | 2,109 | 894 | -58% | 0 | 0 | — |
case-15 | pass→pass | 7,943 | 6,092 | -23% | 1 | 1 | 0% | 1,442 | 1,538 | +7% | 0 | 0 | — |
case-16 | fail→pass | 4,151 | 3,084 | -26% | 1 | 1 | 0% | 655 | 953 | +45% | 0 | 0 | — |
case-17 | pass→pass | 7,572 | 4,990 | -34% | 1 | 1 | 0% | 1,231 | 1,310 | +6% | 0 | 0 | — |
case-18 | pass→pass | 12,658 | 13,468 | +6% | 1 | 1 | 0% | 2,083 | 2,658 | +28% | 0 | 0 | — |
case-19 | pass→pass | 4,625 | 2,662 | -42% | 1 | 1 | 0% | 760 | 922 | +21% | 0 | 0 | — |
case-20 | pass→pass | 9,169 | 8,111 | -12% | 1 | 1 | 0% | 1,897 | 2,133 | +12% | 0 | 0 | — |
case-21 | pass→pass | 9,625 | 7,970 | -17% | 1 | 1 | 0% | 1,785 | 1,990 | +11% | 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 +9 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.