Skill-cleaner

扫描工作区所有 Skill，检测授权、配置和环境变量是否齐全， 并可按需禁用或卸载无效 Skill。

This skill appears to implement the advertised feature set, but review before use: - It calls the 'openclaw' command-line tool, yet the skill declares no required binaries; ensure the 'openclaw' CLI is installed and works on the agent host or the script will fail or report nothing. - The script reads ~/.openclaw/openclaw.json and the workspace skills directory to determine disabled/installed skills; make sure you trust the environment and that these paths are correct for your setup. - The code checks for environment variable names found in other skills' manifests (to detect missing API keys) — it only checks presence and does not exfiltrate values, but it will report which keys are missing. If you have sensitive naming conventions you do not want enumerated, be aware of this behavior. - The 'clean' action runs 'openclaw skills disable' or 'openclaw skills uninstall --force' — these are destructive operations. Always run 'report' first, inspect the list, and preferably run the cleanup in a controlled environment or with a backup. - The implementation contains a bug: diagnose_skill references SkillStatus.CONFIG_INVALID which is not defined in SkillStatus; this can cause runtime exceptions when a manifest is missing or invalid. Consider fixing or testing the script yourself before relying on automated cleanup. If you decide to install: run the script manually first (python3 skill_cleaner.py) to verify output, inspect the code locally, and avoid running the 'clean' action until you confirm the report is accurate.

Type: OpenClaw Skill Name: skill2cleaner Version: 1.0.0 The skill bundle is a management utility designed to audit and maintain the health of installed OpenClaw skills. The Python script (skill_cleaner.py) scans the local workspace, checks for missing environment variables defined in skill manifests, and provides options to disable or uninstall invalid skills using the 'openclaw' CLI. The code uses safe subprocess execution (passing arguments as lists) and limits its file operations to the standard OpenClaw configuration and skill directories, showing no signs of malicious intent or data exfiltration.