The Agent Skills Directory

[SAFE]: The skill provides instructions for generating concise session summaries to help with context continuity without accessing unrelated system files.
[SAFE]: Includes explicit security and privacy constraints, directing the agent to redact secrets, credentials, and personal information from summaries.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data including conversation history and command results (ingested via conversation context). It defines clear boundaries in SKILL.md by limiting the scope to the active session and providing redaction instructions (sanitization). The lack of tool-based execution reduces the risk of following instructions embedded in processed data.

summarize