Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →撰写 Xiaomi Album Syncer 的 Release Notes。适用于：(1) 发版前为指定版本号生成更新日志；(2) 确定版本的变更范围（前序 tag 比较基准）；(3) 校对已有 release notes 是否符合书写规范。
.claude/skills/coooolfan-release-note-creator/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -55% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -47% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -44% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -34% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -39% | 0% |
references/range-determination.md，了解如何确定变更范围references/release-note-guideline.md，了解如何撰写更新日志docs/release_notes/<version>.md.github/workflows/cd.yml）通过 workflow_dispatch 手动触发，create-draft-release.sh 会在目标 commit 上检查 docs/release_notes/<version>.md，文件缺失会导致发布失败prepare-release.sh 校验）：0.17.0 或 0.17.0-rc.1，预发布标签仅允许 alpha、beta、rc，推荐使用点（.）分隔序号0.15.2-beta-1（连字符分隔），新版本不再使用v 前缀gh 命令，仅基于本地 git 信息工作scripts/ 下的脚本都支持传递 --help 参数查看使用方法| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,370 | 8,804 | +38% | 1 | 1 | 0% | 325 | 683 | +110% | 0 | 0 | — |
case-02 | fail→fail | 5,350 | 4,902 | -8% | 1 | 1 | 0% | 294 | 679 | +131% | 0 | 0 | — |
case-03 | fail→fail | 5,134 | 5,335 | +4% | 1 | 1 | 0% | 327 | 713 | +118% | 0 | 0 | — |
case-04 | fail→fail | 8,337 | 14,701 | +76% | 1 | 1 | 0% | 1,658 | 2,264 | +37% | 0 | 0 | — |
case-05 | fail→fail | 2,658 | 7,545 | +184% | 1 | 1 | 0% | 500 | 772 | +54% | 0 | 0 | — |
case-06 | fail→fail | 9,908 | 4,191 | -58% | 1 | 1 | 0% | 1,793 | 1,123 | -37% | 0 | 0 | — |
case-07 | fail→pass | 10,316 | 2,931 | -72% | 1 | 1 | 0% | 1,852 | 835 | -55% | 0 | 0 | — |
case-08 | fail→pass | 11,793 | 3,703 | -69% | 1 | 1 | 0% | 1,922 | 1,016 | -47% | 0 | 0 | — |
case-09 | fail→pass | 9,309 | 2,721 | -71% | 1 | 1 | 0% | 1,659 | 922 | -44% | 0 | 0 | — |
case-10 | fail→pass | 11,256 | 5,164 | -54% | 1 | 1 | 0% | 1,988 | 1,315 | -34% | 0 | 0 | — |
case-11 | pass→pass | 7,835 | 3,293 | -58% | 1 | 1 | 0% | 1,420 | 1,031 | -27% | 0 | 0 | — |
case-12 | fail→fail | 8,777 | 3,022 | -66% | 1 | 1 | 0% | 1,592 | 756 | -53% | 0 | 0 | — |
case-13 | fail→pass | 9,187 | 2,852 | -69% | 1 | 1 | 0% | 1,522 | 925 | -39% | 0 | 0 | — |
case-14 | pass→pass | 7,649 | 3,340 | -56% | 1 | 1 | 0% | 1,335 | 917 | -31% | 0 | 0 | — |
case-15 | fail→pass | 10,166 | 4,321 | -57% | 1 | 1 | 0% | 1,594 | 1,159 | -27% | 0 | 0 | — |
case-16 | pass→pass | 10,567 | 2,384 | -77% | 1 | 1 | 0% | 1,827 | 714 | -61% | 0 | 0 | — |
case-17 | pass→fail | 5,994 | 29,457 | +391% | 1 | 1 | 0% | 966 | 717 | -26% | 0 | 0 | — |
case-18 | fail→fail | 9,273 | 4,497 | -52% | 1 | 1 | 0% | 1,547 | 613 | -60% | 0 | 0 | — |
case-19 | fail→pass | 14,921 | 3,671 | -75% | 1 | 1 | 0% | 2,836 | 907 | -68% | 0 | 0 | — |
case-20 | fail→fail | 9,721 | 4,241 | -56% | 1 | 1 | 0% | 1,536 | 1,097 | -29% | 0 | 0 | — |
case-21 | fail→fail | 8,014 | 1,297 | -84% | 1 | 1 | 0% | 1,273 | 560 | -56% | 0 | 0 | — |
case-22 | fail→pass | 9,020 | 2,587 | -71% | 1 | 1 | 0% | 1,563 | 754 | -52% | 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 16 counted toward the lift figure. The other 6 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 +32 percentage points is the difference between those two pass rates over the 16 comparable cases. 3 cases got worse with the skill loaded, and they are 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.