The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes the entire conversation history to generate intellectual and technical summaries. Because this chat history is untrusted external data, it presents an attack surface where embedded instructions could attempt to influence the agent's output or actions. The instructions lack explicit boundary markers or sanitization steps to differentiate between the user's conversation content and the archival instructions.
[COMMAND_EXECUTION]: The skill performs file system operations by writing archival notes to specific paths. It supports using a provided save-location or automatically detecting relevant folders in the user's environment to store the output. This capability to write to the file system based on potentially untrusted metadata or folder structures is a functional requirement but remains a point of observation for security auditing.

archive-conversation