mac-cleanup

This skill's stated purpose (safe macOS cleanup) matches the commands and flows provided: it scans filesystem locations, groups candidates by risk, asks for explicit user confirmation, and executes destructive cleanup commands. The primary security concern is the heavy use of destructive shell commands (rm -rf, docker prune) which are inherently high-impact. Safety mitigations are present in the document (explicit confirmation required, show sizes before deleting, never delete paths in safe-cleanup-paths.md), but the fragment places trust on external files (safe-cleanup-paths.md) and user/operator vigilance. The use of 2>/dev/null across commands reduces visibility and auditing. There are no explicit network exfiltration or remote download patterns in the skill text, and no hardcoded credentials or obfuscated code. Overall, the skill is not malicious by content, but it is high-risk operationally: if implemented or used carelessly (blind approvals, manipulated safe path lists, or malicious local tools), it can cause significant data loss. Recommend enforcing programmatic canonicalization checks for any target path, avoiding broad rm -rf patterns without absolute-path verification, removing excessive stderr suppression for transparency, and requiring granular per-item confirmations for medium-risk deletes.