The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local shell commands like grep and sed to interact with the filesystem. These commands are used to locate and manage the agent's internal memory files within the user's home directory.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it processes MEMORY.md, which contains historical data that could be influenced by external inputs in previous sessions.
Ingestion points: Reads from MEMORY.md in Step 2.
Boundary markers: The skill includes a manual confirmation step where the user must verify the pattern found in memory before it is processed.
Capability inventory: The skill has the ability to read and write to local configuration files (CLAUDE.md, .claude/rules/) and delete entries from MEMORY.md via shell commands.
Sanitization: The distillation process in Step 4 acts as a logical filter, and the final confirmation step in Step 7 ensures the user reviews the final instruction before it is permanently enforced.

promote