Skill Sync

Audit and converge local skills across Codex, Claude, OpenClaw, OpenCode, workspace skills, and shared agent libraries; compute a hygiene score, classify sha...

This project appears to do what it claims: scanning local agent skill directories, producing a hygiene report, and optionally converging duplicates by moving originals into backups and creating symlinks. Before running any destructive action: 1) Run scans and previews only (omit --apply) to inspect recommended operations. 2) Verify the backup location (default ~/.skill-sync/backups) and/or set SKILL_SYNC_BACKUP_ROOT to a safe path. 3) Inspect the export manifest (--export-manifest) and review operations the import preview reports. 4) Avoid using install.sh --force unless you understand it will remove/replace destinations. 5) Test in a disposable directory or VM (the included tests are a good model) if you have many important skills. If you need extra assurance, review the full scripts/skill_sync.py to confirm the exact move/backup logic and run the unit tests locally.

Type: OpenClaw Skill Name: multi-agent-skill-sync Version: 1.2.1 The skill bundle provides a utility for managing, deduplicating, and syncing AI agent skills across various platforms (Codex, Claude, OpenClaw, etc.). The core logic in `scripts/skill_sync.py` performs file system operations such as scanning skill directories, calculating SHA256 hashes to identify duplicates, and replacing redundant copies with symlinks to a canonical source. While the script has broad write access to several agent-related directories in the user's home folder, it includes a robust backup and restore mechanism located in `~/.skill-sync/backups` and lacks any indicators of data exfiltration, network activity, or unauthorized persistence. The behavior is consistent with the stated purpose in `SKILL.md` and `README.md`.