Install
$ agentstack add skill-runesleo-claude-skill-audit-claude-skill-audit ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
Skill Audit — 技能树体检
> 对照 Tw93 Waza 的 /health 维度。56+ skill 没人管就是 tech debt。
触发场景
- 周日 /today 自动建议
- 手动 /skill-audit
- 新增 >5 个 skill 后复盘
- 半年一次基线对比
执行流程
Step 1 — 清点家底
ls ~/.claude/skills/ | grep -v '^_\|\.md$\|\.jsonl$' | wc -l
ls ~/.claude/skills/ | grep -v '^_\|\.md$\|\.jsonl$' > /tmp/skill-list.txt
Step 2 — 热度分析(execution-log.jsonl)
cat ~/.claude/skills/execution-log.jsonl | jq -r '.skill' | sort | uniq -c | sort -rn
注意:log 可能严重欠记录(当前 516B)。同步查 patterns.md / today.md 最近 30 天触发痕迹:
grep -roh '[a-z-]*-skill\|/[a-z-]*' ~/.claude/memory/today.md ~/.claude/memory/yesterday.md 2>/dev/null | sort -u
Step 3 — 死 skill 识别
连续 30 天 0 触发 + 无 hook 绑定 + 无 CLAUDE.md 引用 → 候选死 skill。
for skill in $(cat /tmp/skill-list.txt); do
cnt=$(grep -c "\"skill\":\"$skill\"" ~/.claude/skills/execution-log.jsonl 2>/dev/null || echo 0)
ref=$(grep -r "$skill" ~/.claude/rules/ ~/.claude/CLAUDE.md 2>/dev/null | wc -l)
echo "$skill | log:$cnt | ref:$ref"
done | sort -t'|' -k2n
Step 4 — 冲突检测
两个 skill 触发条件高度重叠 = 冲突。扫 description 字段关键词:
- review / codex review
- debug / systematic-debugging
- xhs-publish / xhs-style / xhs-review
- content-pipeline / article-pipeline
输出:"⚠️ [A] 和 [B] 可能重叠:共同触发词 = [...]"
Step 5 — 过期 pattern 清理
# patterns.md 连续 5 session 未触发的条目
grep -n "pattern_" ~/.claude/skills/leo-evolution/learned/patterns.md 2>/dev/null | head -20
建议用户手动 review,不自动删。
Step 6 — 输出报告
格式:
📊 Skill Tree Audit — YYYY-MM-DD
📈 家底:56 skill / 4 hook / 12 别名
🔥 Top 5 热度:leo-style(12) / today(8) / recall(6) / codex(4) / x-research(3)
💀 死 skill 候选:[列表](建议移到 _archived/)
⚠️ 冲突疑似:[列表]
🧹 过期 pattern:[数量],见 patterns.md 第 N 行
🎯 行动项:
- [ ] 归档死 skill
- [ ] 合并冲突 skill
- [ ] 清理过期 pattern
保存基线:报告写入 ~/.claude/memory/skill-tree-audit-YYYY-MM-DD.md,下次对比增量。
设计哲学
- 不自动删:skill 是肌肉记忆,死 skill 也可能是季节性触发(/tweet-review 两周一次)
- 归档 > 删除:候选死 skill 移到
_archived/,保留 30 天再永久删 - 人工 review:自动识别 + 人工判断,不做全自动清理
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: runesleo
- Source: runesleo/claude-skill-audit
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.