The Agent Skills Directory

[COMMAND_EXECUTION]: The skill employs the letta CLI tool and standard git commands (git status, git add, git commit, git push) to manage the agent's internal memory state. These commands are used as intended for synchronizing and maintaining the agent's persistent storage.
[DATA_EXFILTRATION]: While the skill performs a git push of the memory directory, this action is consistent with the Letta platform's architecture for persisting agent state. The operation targets the agent's own memory repository rather than an unauthorized external service.
[PROMPT_INJECTION]: The skill contains instructions for the agent to modify its own system prompt. However, these are accompanied by clear safety guidelines requiring conservative edits, preservation of the user's identity, and maintenance of the agent's core persona.
[DATA_EXPOSURE]: The skill provides the agent with a methodology for scanning its own memory files for redundancy and bloat. This is a self-referential management task and does not involve exposing sensitive system credentials or external user data.

Context Doctor