The Agent Skills Directory

[SAFE]: The skill implements proactive security measures, including path normalization to prevent directory traversal (..) and a redaction scan to identify and mask secrets or credentials before processing content.
[SAFE]: The skill uses mcp__codex__codex for verification. This tool is a vendor-owned resource for this skill author, and the implementation uses a read-only sandbox with an approval-policy: 'never' to ensure secure, non-modifying execution.
[SAFE]: Data exfiltration is mitigated by strict path validation enforcing repository boundaries (git rev-parse --show-toplevel) and a local-first writing strategy.

fp-brief