The Agent Skills Directory

[SAFE]: The skill operates under a strict 'no-write' policy for production code, commits, and pull requests. Its activity is limited to internal reasoning, dialogue with the user, and writing session notes to a specific directory (.cheese/notes/).
[SAFE]: Code analysis tools (cheez-search, tilth_search) are used exclusively for read-only 'shape checks' to determine the scope of potential changes, which is consistent with the skill's purpose as a reasoning tool.
[SAFE]: The handoff mechanism provides a structured way to pass reasoning outcomes to other internal agent skills. While it processes user input to generate these summaries, it does not involve any unsafe execution or privilege escalation vectors.

culture