Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Turn changelog-scan output into polished, categorized release notes draft. Propose only.
.claude/skills/cobusgreyling-draft-release-notes/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -4% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 136% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -14% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 73% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 39% | 0% |
changelog-scan (the list of items + summary)Write a ready-to-review draft to RELEASE_NOTES_DRAFT.md (or print it clearly so the loop can save it).
Use this structure (adapt section names to what actually exists; omit empty sections):
markdown# Release Notes — vX.Y.Z (unreleased) ## Breaking Changes - ... ## Features - ... ## Bug Fixes - ... ## Performance - ... ## Security - ... ## Documentation & Examples - ... ## Internal / Maintenance (usually omitted from public notes) - ... **Thanks** to @contributor1, @contributor2 for contributions to this release. **Full changelog**: https://github.com/ORG/REPO/compare/vPREV...HEAD
After writing the draft, the loop should update state with the draft location and "pending human review".
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 7,129 | 4,351 | -39% | 1 | 1 | 0% | 1,353 | 1,301 | -4% | 0 | 0 | — |
case-02 | fail→pass | 17,351 | 8,414 | -52% | 1 | 1 | 0% | 888 | 2,100 | +136% | 0 | 0 | — |
case-03 | fail→pass | 10,661 | 6,370 | -40% | 1 | 1 | 0% | 2,007 | 1,733 | -14% | 0 | 0 | — |
case-04 | fail→pass | 3,962 | 3,645 | -8% | 1 | 1 | 0% | 607 | 1,052 | +73% | 0 | 0 | — |
case-05 | fail→pass | 5,989 | 5,687 | -5% | 1 | 1 | 0% | 1,042 | 1,446 | +39% | 0 | 0 | — |
case-06 | pass→pass | 5,026 | 4,211 | -16% | 1 | 1 | 0% | 969 | 1,235 | +27% | 0 | 0 | — |
case-07 | pass→pass | 5,453 | 3,196 | -41% | 1 | 1 | 0% | 1,049 | 1,148 | +9% | 0 | 0 | — |
case-08 | fail→pass | 8,342 | 3,292 | -61% | 1 | 1 | 0% | 1,542 | 1,135 | -26% | 0 | 0 | — |
case-09 | pass→pass | 9,040 | 3,569 | -61% | 1 | 1 | 0% | 1,571 | 1,167 | -26% | 0 | 0 | — |
case-10 | pass→pass | 6,604 | 3,244 | -51% | 1 | 1 | 0% | 1,302 | 1,133 | -13% | 0 | 0 | — |
case-11 | pass→pass | 6,483 | 7,284 | +12% | 1 | 1 | 0% | 1,249 | 1,756 | +41% | 0 | 0 | — |
case-12 | pass→pass | 3,939 | 3,248 | -18% | 1 | 1 | 0% | 748 | 1,177 | +57% | 0 | 0 | — |
case-13 | pass→pass | 6,490 | 5,424 | -16% | 1 | 1 | 0% | 1,167 | 1,524 | +31% | 0 | 0 | — |
case-14 | pass→pass | 5,037 | 3,696 | -27% | 1 | 1 | 0% | 914 | 1,216 | +33% | 0 | 0 | — |
case-15 | pass→pass | 6,339 | 3,883 | -39% | 1 | 1 | 0% | 1,133 | 1,131 | -0% | 0 | 0 | — |
case-16 | fail→pass | 5,944 | 3,559 | -40% | 1 | 1 | 0% | 1,054 | 1,147 | +9% | 0 | 0 | — |
case-17 | fail→pass | 6,732 | 3,593 | -47% | 1 | 1 | 0% | 1,246 | 1,169 | -6% | 0 | 0 | — |
case-18 | fail→pass | 7,866 | 4,561 | -42% | 1 | 1 | 0% | 1,426 | 1,339 | -6% | 0 | 0 | — |
case-19 | fail→pass | 7,626 | 3,487 | -54% | 1 | 1 | 0% | 1,437 | 1,044 | -27% | 0 | 0 | — |
case-20 | pass→pass | 7,227 | 9,158 | +27% | 1 | 1 | 0% | 1,430 | 2,290 | +60% | 0 | 0 | — |
case-21 | pass→pass | 5,290 | 5,401 | +2% | 1 | 1 | 0% | 932 | 1,435 | +54% | 0 | 0 | — |
case-22 | pass→pass | 5,287 | 7,072 | +34% | 1 | 1 | 0% | 1,071 | 1,928 | +80% | 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 +45 percentage points is the difference between those two pass rates over the 21 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.