The Agent Skills Directory

[SAFE]: The skill operates entirely locally on a provided context_root path. It is designed to read markdown and JSON files to identify stale records (like overdue portfolio reviews or dormant learning logs) without making any external connections.
[SAFE]: The Python implementation uses pathlib for file system traversal and re for parsing, avoiding any eval(), exec(), or subprocess spawning.
[SAFE]: Frontmatter configuration properly restricts the skill to be non-user-invocable, intended for internal agent orchestration rather than direct user interaction.
[SAFE]: Sensitive data exposure is avoided by focusing on metadata (dates, status lines) rather than full content extraction, and the tool specifically declares itself as read-only.

cadence-check