The Agent Skills Directory

[SAFE]: The skill is designed with significant safety guardrails. It processes untrusted external data (meeting transcripts) but limits its direct actions to writing new files in a specific meetings/ directory using collision-safe naming conventions. Crucially, it prohibits the automatic update of core project files like PROJECT.md or TASKS.md, requiring the user to manually select specific items for 'promotion' to other skills. This architecture maintains a clear boundary between potentially malicious external input and the project's durable state.

meeting-minutes