skill-registry-scan
Skill1 starsChanged 7 months agoMITView on GitHub
小能核心技能 — 扫描所有workspace的skills/目录,自动生成/更新skill-registry.json。多信号成熟度推断+异常检测+批量模式。触发词:扫描registry/更新技能目录/小能巡检/scan skills
In this repository6 other files it gives its agents.
- 小能Skill
- agent-capability-mapSkill
- skill-auditSkill
- skill-card-genSkill
- skill-forgeSkill
- skill-graphSkill
The file
Files
Skill Registry Scan
v2.0 | 升级:多信号成熟度推断+异常检测+批量扫描
触发词
- "扫描 registry" / "更新技能目录" / "小能巡检" / "scan skills"
流程
Step 1:扫描目录
扫描范围(默认当前workspace,--all扫描所有workspace):
/root/.openclaw/workspace/skills/*/SKILL.md/root/.openclaw/workspace-lengjing/skills/*/SKILL.md/root/.openclaw/workspace-xiaokai/skills/*/SKILL.md/root/.openclaw/shared-skills/*/SKILL.md
对每个SKILL.md提取:
name(frontmatter)description(frontmatter)version(如有)- 是否有
references/目录 - 是否有
memory/目录 - 是否有
scripts/目录 - 是否有
evals/目录 - SKILL.md 行数
- frontmatter完整性(name+description是否齐全)
Step 2:对比现有 registry
读取 memory/skill-registry.json,对比:
- 新增:目录中有但registry中没有的
- 修改:SKILL.md最后修改时间比registry的last_scan新的
- 删除:registry中有但目录中已不存在的
Step 3:多信号成熟度推断
不再只看行数,综合5个信号:
| 信号 | 权重 | 评分规则 |
| SKILL.md行数 | 20% | >500行=5, 300-500=4, 200-300=3, 100-200=2, <100=1 |
| references数量 | 25% | ≥3个=5, 2个=4, 1个=3, 0个=1 |
| memory数量 | 15% | ≥5个=5, 3-4个=4, 1-2个=3, 0个=1 |
| frontmatter完整性 | 20% | name+description+version=5, name+description=4, 只有name=2, 缺失=1 |
| 辅助目录 | 20% | scripts+evals都有=5, 有其一=3, 都没有=1 |
成熟度映射:
- 综合分4.5-5.0 → L4(生产级)
- 综合分3.5-4.4 → L3(可用)
- 综合分2.5-3.4 → L2(实验)
- 综合分1.0-2.4 → L1(概念)
Step 4:异常检测
自动标记以下异常:
| 异常类型 | 检测规则 | 严重度 |
| 空SKILL.md | 行数<10或内容为空 | 🔴 严重 |
| frontmatter缺失 | 没有name或description | 🔴 严重 |
| description过短 | description<20字符 | 🟡 警告 |
| 无references的大型Skill | >300行且无references/ | 🟡 警告 |
| 无触发词 | description中无明确触发场景 | 🟡 警告 |
| 孤立Skill | 与其他Skill无任何关系 | 🟢 提示 |
Step 5:更新 registry.json
增量更新 memory/skill-registry.json:
{
"name": "skill-name",
"path": "skills/skill-name/",
"agent": "小能",
"version": "v2.0",
"status": "active",
"maturity": "L3",
"maturity_signals": {
"lines": 115,
"refs_count": 2,
"mem_count": 1,
"frontmatter": "complete",
"aux_dirs": ["scripts"]
},
"maturity_score": 4.2,
"type_primary": "管理类",
"anomalies": [],
"last_scan": "2026-05-01T16:00:00+08:00"
}Step 6:输出变更摘要
📊 Skill Registry 扫描完成
━━━━━━━━━━━━━━━━━━━━
扫描范围:{workspace-list}
总计:XX 个 Skill(active XX | deprecated XX)
🆕 新增:X 个
- skill-name — description [L2]
✏️ 修改:X 个
- skill-name — 变更说明
🗑️ 删除:X 个
- skill-name
⚠️ 异常:X 个
- skill-name — 🔴 空SKILL.md
- skill-name — 🟡 无触发词
📈 成熟度分布:L4:{N} | L3:{N} | L2:{N} | L1:{N}输出文件
memory/skill-registry.json— 更新后的完整registrymemory/scan-log.md— 追加本次扫描记录
注意事项
- 只追加/修改,不从头重建registry
- 新Skill自动标记status: "active",成熟度为推断值
- 异常Skill(空SKILL.md/frontmatter缺失)标为"needs_review"
- 扫描结果喂给skill-audit(深度评估)、skill-graph(关系推断)、agent-capability-map(能力地图)
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

