skill-stocktake(技能质量盘点 · DSH 版)
对 DSH 技能库做质量盘点的命令式流程:质量检查表 + 整体 AI 判断。两模式:Quick Scan(仅自上次运行以来变更的技能)与 Full Stocktake(全盘)。
范围与扫描根
盘点目标为 DSH 四根(优先级降序):
| 根 | 说明 |
|---|---|
<项目>/.dsh/skills/ |
项目级技能(目录存在时) |
<项目>/.agents/skills/ |
项目级技能(目录存在时) |
~/.dsh/skills/ |
用户级技能 |
~/.agents/skills/ |
用户级技能(skill_manage 库默认位) |
Phase 1 开工时必须显式列出找到了哪些根、各扫到多少文件(:21 原文纪律:"At the start of Phase 1, the command explicitly lists which paths were found and scanned.")。要盘点特定项目,就在该项目根目录发起;项目无技能目录时只盘用户级(:32 语义对应)。
模式
| 模式 | 触发 | 时长 |
|---|---|---|
| Quick Scan | results.json 存在(缺省) |
5–10 min |
| Full Stocktake | results.json 不存在,或显式 full |
20–30 min |
Results 缓存:<项目>/agent-out/skill-stocktake-results.json(缓存与断点续跑语义保留,C 改造点④)。
Quick Scan 流程
只重评自上次运行以来变更的技能(5–10 min):
- 读 results 缓存;
- 对四根做 mtime diff;
- diff 为空 → 报告「自上次运行以来无变更」并停止;
- 只用 Phase 2 同一标准重评变更件;
- 未变更技能结转上次结果;
- 只输出 diff;
- 把结果写回 results 缓存。
Full Stocktake 流程
Phase 1 — 盘点
扫描四根:枚举技能文件、提取 frontmatter、收集 mtime。呈现扫描摘要与库存表,格式对齐 ECC 原文(找到/未找到逐根标注):
Scanning:
✓ ~/.dsh/skills/ (N files)
✓ ~/.agents/skills/ (N files)
✗ <项目>/.dsh/skills/ (not found — 用户级 only)
库存表——ECC 原表头保留如下(:74-75 原文):
| Skill | 7d use | 30d use | Description |
|-------|--------|---------|-------------|
Phase 2 — 质量评价
启动批评席子会话,携带完整库存与检查表:
- Freshness of technical references verified (use WebSearch if tool names / CLI flags / APIs are present)
- Usage frequency considered
裁决判据:
| Verdict | Meaning |
|---------|---------|
| Keep | Useful and current |
| Improve | Worth keeping, but specific improvements needed |
| Update | Referenced technology is outdated (verify with WebSearch) |
| Retire | Low quality, stale, or cost-asymmetric |
| Merge into [X] | Substantial overlap with another skill; name the merge target |
评价是**整体 AI 判断**,不是数值打分(:126 原文纪律)。导向维度四条(:126-130 保留):Actionability(有能立即行动的代码/命令/步骤)、Scope fit(名称/触发/内容对齐,不宽不窄)、Uniqueness(价值不可被 AGENTS.md/记忆/其他技能替代)、Currency(技术引用在当前环境可用)。
**判定理由质量要求(承重段,勿改写)**:
Reason quality requirements — the reason field must be self-contained and decision-enabling:
- Do NOT write "unchanged" alone — always restate the core evidence
- For Retire: state (1) what specific defect was found, (2) what covers the same need instead
- Bad:
"Superseded" - Good:
"disable-model-invocation: true already set; superseded by continuous-learning-v2 which covers all the same patterns plus confidence scoring. No unique content remains." - For Merge: name the target and describe what content to integrate
- Bad:
"Overlaps with X" - Good:
"42-line thin content; Step 4 of chatlog-to-article already covers the same workflow. Integrate the 'article angle' tip as a note in that skill." - For Improve: describe the specific change needed (what section, what action, target size if relevant)
- Bad:
"Too long" - Good:
"276 lines; Section 'Framework Comparison' (L80–140) duplicates ai-era-architecture-principles; delete it to reach ~150 lines." - For Keep (mtime-only change in Quick Scan): restate the original verdict rationale, do not write "unchanged"
- Bad:
"Unchanged" - Good:
"mtime updated but content unchanged. Unique Python reference explicitly imported by rules/python/; no overlap found."
### Phase 3 — 汇总表
ECC 原表头保留(:149):
```markdown
| Skill | 7d use | Verdict | Reason |
|-------|--------|---------|--------|
Phase 4 — 处置
- Retire / Merge:逐文件给出详尽论证后再向用户确认:发现了什么具体问题(重叠/过时/引用失效等);什么替代物覆盖同一需求(Retire 指到既有技能/规则,Merge 指到目标文件与要整合的内容);移除的影响面(依赖技能、AGENTS.md/记忆引用、受影响工作流)。
- Improve:给出具体改进建议与理由(如「430→200 行,因 X/Y 节与某技能重复」);用户决定是否执行。
- Update:给出核验过来源的更新内容。 4.检查 AGENTS.md/记忆轨规模,超限提议压缩。
归档/删除操作永远需要用户显式确认(:194 原文纪律)。
Results 文件模式
evaluated_at 必须是评价完成时刻的真实 UTC 时间(ECC :168-169 原文纪律:经命令取 date -u +%Y-%m-%dT%H:%M:%SZ 形态,不得用 T00:00:00Z 一类日期近似值充数。):
{
"evaluated_at": "2026-02-21T10:00:00Z",
"mode": "full",
"batch_progress": {
"total": 80,
"evaluated": 80,
"status": "completed"
},
"skills": {
"skill-name": {
"path": "~/.agents/skills/skill-name/SKILL.md",
"verdict": "Keep",
"reason": "Concrete, actionable, unique value for X workflow",
"mtime": "2026-01-15T08:30:00Z"
}
}
}
Notes
- 评价是盲评:同一检查表适用于所有技能,不论来源(ECC、自研、自动提取)
- 归档/删除操作永远需要用户显式确认
- 不按技能来源分支裁决