360 Secure SkillHub
360 安全技能中心

Skills Audit

该 Skill 运行 Python 脚本扫描本地技能目录中的 SKILL.md，依据金名单和平台关键词将技能分类为已批准、待复核或平台专属，生成审计报告，并在用户确认后删除冗余技能目录。

anderszhan
anderszhan
热度 16.02
最近更新 2026-06-17 23:16:34
分类：效率依赖：代码执行 / 读写文件应用场景：工作流 / 报告撰写

安装技能

复制下面的话术发给你的 Agent，它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill： npx secure-skills-cli add workbuddy-skills-audit
Metadata
name
skills-audit
description
Audit all installed skills to identify duplicates, platform-mismatched skills, and maintenance candidates. Use when: (1) the user asks to "clean up skills", "check for duplicate skills", "audit installed skills", or "find skills to remove"; (2) performing periodic skill maintenance; (3) investigating skill clutter or skill conflicts. Scans all SKILL.md files in ~/.workbuddy/skills/, compares against a golden list of approved skills, classifies each as approved / unknown / platform-specific, and generates a structured audit report.

Skills Audit

Audit all installed WorkBuddy skills, classify them, and generate a cleanup report.

Workflow

1. Run the audit script

python scripts/audit_skills.py

This scans ~/.workbuddy/skills/, reads each SKILL.md, and generates a skills-audit-YYYY-MM.md report.

Options:

  • --dir <path> — custom skills directory (default: ~/.workbuddy/skills/)
  • --golden <path> — custom golden list (default: references/golden-list.md)
  • --output <path> — custom report output path

2. Interpret the report

The report classifies each skill into one of three categories:

Status Meaning Action
[OK] Approved Matches the golden list Keep
[!] Review Not in golden list, not platform-specific Manually review
[X] Platform: Designed for Clawdbot/Moltbot/OpenClaw Candidate for removal

3. Present findings to user

Use present_files to show the report. For platform-specific or unknown skills, flag them and ask the user for confirmation before removing.

4. Remove redundant skills (after user confirmation)

rm -rf ~/.workbuddy/skills/<skill-dir>

Classification Rules

  1. Golden list match: If the skill name or directory matches an entry in references/golden-list.md, it is Approved.
  2. Platform detection: SKILL.md is scanned for keywords (clawdbot, moltbot, openclaw, clawhub, CodeConductor.ai). If found and not in golden list, it is flagged as platform-specific.
  3. Otherwise: Classified as Unknown — requires manual review.

Bundled Resources

  • scripts/audit_skills.py — Python script that performs the scan and generates the report
  • references/golden-list.md — Canonical list of approved skills (edit this to add/remove approved skills)