Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Auto-generate CHANGELOG.md entries from git commit history. Activate when user mentions "changelog", "release notes", "what changed", "generate changelog", "CHANGELOG", "release history".
.claude/skills/bilal140202-doncheli-changelog/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 122% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -38% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -54% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -63% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -33% | 0% |
--from/--to: use the specified refs--version: use that as the new version numbergit log --oneline <range> and parse conventional commit prefixesfeat / feat! → Added (breaking changes go in a separate top section)fix → Fixedrefactor / perf → Changeddocs → Changedsecurity → Securitydeprecated → Deprecatedremove → Removedtest / chore / ci → omit by default (include with --verbose)Merge branch, Merge pull request)CHANGELOG.md, below the header — never overwrite existing entriesCHANGELOG.md exists, create it with the standard Keep a Changelog header--dry-run, print to stdout only — do not write to filegit add CHANGELOG.mdfeat!, BREAKING CHANGE footer), highlight them prominently--version| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 3,477 | 2,274 | -35% | 1 | 1 | 0% | 456 | 649 | +42% | 0 | 0 | — |
case-02 | fail→pass | 6,602 | 15,068 | +128% | 1 | 1 | 0% | 1,281 | 2,849 | +122% | 0 | 0 | — |
case-03 | fail→pass | 12,262 | 5,252 | -57% | 1 | 1 | 0% | 2,298 | 1,429 | -38% | 0 | 0 | — |
case-08 | pass→pass | 5,439 | 2,819 | -48% | 1 | 1 | 0% | 1,113 | 984 | -12% | 0 | 0 | — |
case-04 | pass→pass | 9,755 | 4,040 | -59% | 1 | 1 | 0% | 1,769 | 1,069 | -40% | 0 | 0 | — |
case-05 | fail→fail | 9,411 | 3,795 | -60% | 1 | 1 | 0% | 2,194 | 1,143 | -48% | 0 | 0 | — |
case-06 | pass→pass | 9,332 | 2,728 | -71% | 1 | 1 | 0% | 1,686 | 865 | -49% | 0 | 0 | — |
case-07 | pass→pass | 5,521 | 3,649 | -34% | 1 | 1 | 0% | 1,143 | 1,055 | -8% | 0 | 0 | — |
case-09 | fail→pass | 8,843 | 1,978 | -78% | 1 | 1 | 0% | 1,516 | 693 | -54% | 0 | 0 | — |
case-10 | fail→pass | 9,029 | 2,988 | -67% | 1 | 1 | 0% | 1,917 | 715 | -63% | 0 | 0 | — |
case-11 | pass→pass | 4,803 | 3,681 | -23% | 1 | 1 | 0% | 962 | 839 | -13% | 0 | 0 | — |
case-12 | pass→pass | 3,457 | 1,933 | -44% | 1 | 1 | 0% | 600 | 696 | +16% | 0 | 0 | — |
case-13 | fail→fail | 9,940 | 3,712 | -63% | 1 | 1 | 0% | 1,846 | 1,063 | -42% | 0 | 0 | — |
case-14 | pass→pass | 7,629 | 2,271 | -70% | 1 | 1 | 0% | 1,542 | 815 | -47% | 0 | 0 | — |
case-15 | fail→pass | 8,554 | 3,600 | -58% | 1 | 1 | 0% | 1,664 | 1,111 | -33% | 0 | 0 | — |
case-16 | pass→pass | 4,011 | 1,325 | -67% | 1 | 1 | 0% | 694 | 599 | -14% | 0 | 0 | — |
case-17 | fail→pass | 9,375 | 4,980 | -47% | 1 | 1 | 0% | 1,786 | 1,364 | -24% | 0 | 0 | — |
case-18 | fail→pass | 8,212 | 1,479 | -82% | 1 | 1 | 0% | 1,446 | 614 | -58% | 0 | 0 | — |
case-19 | pass→pass | 4,537 | 3,101 | -32% | 1 | 1 | 0% | 851 | 1,034 | +22% | 0 | 0 | — |
case-20 | pass→pass | 1,264 | 1,319 | +4% | 1 | 1 | 0% | 198 | 554 | +180% | 0 | 0 | — |
case-21 | pass→pass | 2,887 | 2,497 | -14% | 1 | 1 | 0% | 530 | 796 | +50% | 0 | 0 | — |
case-22 | pass→pass | 4,972 | 2,541 | -49% | 1 | 1 | 0% | 1,015 | 882 | -13% | 0 | 0 | — |
case-23 | pass→pass | 5,711 | 3,856 | -32% | 1 | 1 | 0% | 1,119 | 1,123 | +0% | 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. 23 cases were attempted. The headline lift of +30 percentage points is the difference between those two pass rates over the 23 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.