The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a bundled Python script session-turns.py to analyze the current session's transcript. This is a local operation intended to provide the agent with a complete view of user feedback.
[DATA_EXPOSURE]: The skill reads from and writes to specific configuration files such as CLAUDE.md and MEMORY.md. This is the primary purpose of the skill (persistent memory) and is handled with explicit instructions for user confirmation when conflict or deletion is involved.
[INDIRECT_PROMPT_INJECTION]: The skill processes conversation history which could contain instructions. However, it incorporates a rigorous 'admission test' (Phase 1.6) that requires rules to be testable, recur, and have changed an action, which serves as a functional filter against transient or malicious inputs.

close