Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →ECCのための証拠優先のリポジトリ実行ワークフロー。ユーザーがコマンドの実行、リポジトリの確認、CIの失敗のデバッグ、正確な実行と検証の証明を伴う狭い修正のプッシュを必要とする場合に使用する。
.claude/skills/affaan-m-terminal-ops/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-22 | ✗→✓ | ▲ Improved | 567% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 487% | 0% |
| case-21 | ✗→✓ | ▲ Improved | -32% | 0% |
| case-09 | ✓→✗ | ▼ Worse | -4% | 0% |
| case-11 | ✓→✗ | ▼ Worse | 77% | 0% |
当用户需要真实的仓库执行时使用此技能：运行命令、检查 git 状态、调试 CI 或构建、进行窄幅修复，并准确报告更改和验证的内容。
此技能有意比通用编码指导更窄。它是一种以证据为先的终端执行操作工作流。
在相关时，将这些 ECC 原生技能引入工作流：
verification-loop 用于更改后的精确验证步骤tdd-workflow 当正确的修复需要回归覆盖时security-review 当涉及密钥、认证或外部输入时github-ops 当任务依赖于 CI 运行、PR 状态或发布状态时knowledge-ops 当需要将验证结果捕获到持久的项目上下文中时明确：
在更改任何内容之前：
一次解决一个主要失败：
使用确切的状态词：
text表面 - 仓库 - 分支 - 请求模式 证据 - 失败的命令 / 差异 / 测试 操作 - 变更内容 状态 - 已检查 / 本地已更改 / 本地已验证 / 已提交 / 已推送 / 已阻止
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-22 | fail→pass | 5,929 | 8,279 | +40% | 1 | 1 | 0% | 293 | 1,953 | +567% | 0 | 0 | — |
case-01 | fail→fail | 6,532 | 5,714 | -13% | 1 | 1 | 0% | 252 | 1,062 | +321% | 0 | 0 | — |
case-02 | fail→fail | 5,651 | 5,891 | +4% | 1 | 1 | 0% | 282 | 1,036 | +267% | 0 | 0 | — |
case-03 | fail→fail | 5,264 | 7,531 | +43% | 1 | 1 | 0% | 204 | 1,117 | +448% | 0 | 0 | — |
case-04 | pass→pass | 23,451 | 22,754 | -3% | 1 | 1 | 0% | 3,513 | 4,125 | +17% | 0 | 0 | — |
case-05 | pass→pass | 8,443 | 6,810 | -19% | 1 | 1 | 0% | 1,601 | 1,795 | +12% | 0 | 0 | — |
case-06 | pass→pass | 32,530 | 13,880 | -57% | 1 | 1 | 0% | 2,716 | 3,094 | +14% | 0 | 0 | — |
case-07 | fail→pass | 5,387 | 5,805 | +8% | 1 | 1 | 0% | 253 | 1,484 | +487% | 0 | 0 | — |
case-08 | pass→pass | 5,555 | 5,572 | +0% | 1 | 1 | 0% | 842 | 1,587 | +88% | 0 | 0 | — |
case-09 | pass→fail | 7,412 | 7,380 | -0% | 1 | 1 | 0% | 1,237 | 1,189 | -4% | 0 | 0 | — |
case-10 | fail→fail | 9,341 | 7,373 | -21% | 1 | 1 | 0% | 1,344 | 1,766 | +31% | 0 | 0 | — |
case-11 | pass→fail | 10,264 | 6,436 | -37% | 1 | 1 | 0% | 631 | 1,115 | +77% | 0 | 0 | — |
case-12 | fail→fail | 3,800 | 2,893 | -24% | 1 | 1 | 0% | 595 | 1,177 | +98% | 0 | 0 | — |
case-13 | fail→fail | 14,007 | 10,075 | -28% | 1 | 1 | 0% | 1,642 | 1,071 | -35% | 0 | 0 | — |
case-14 | pass→pass | 9,665 | 4,416 | -54% | 1 | 1 | 0% | 1,537 | 1,489 | -3% | 0 | 0 | — |
case-15 | pass→pass | 7,934 | 14,196 | +79% | 1 | 1 | 0% | 1,254 | 2,468 | +97% | 0 | 0 | — |
case-16 | pass→pass | 11,845 | 5,638 | -52% | 1 | 1 | 0% | 1,820 | 1,635 | -10% | 0 | 0 | — |
case-17 | pass→pass | 9,467 | 3,900 | -59% | 1 | 1 | 0% | 1,494 | 1,333 | -11% | 0 | 0 | — |
case-18 | fail→fail | 21,250 | 13,731 | -35% | 1 | 1 | 0% | 2,900 | 2,962 | +2% | 0 | 0 | — |
case-19 | pass→pass | 14,580 | 10,424 | -29% | 1 | 1 | 0% | 2,367 | 2,310 | -2% | 0 | 0 | — |
case-20 | pass→fail | 11,338 | 7,645 | -33% | 1 | 1 | 0% | 1,838 | 1,132 | -38% | 0 | 0 | — |
case-21 | fail→pass | 17,573 | 6,553 | -63% | 1 | 1 | 0% | 2,683 | 1,816 | -32% | 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 13 counted toward the lift figure. The other 9 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 0 percentage points is the difference between those two pass rates over the 13 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.