The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes contents from internal memory files (e.g., user-profile.md, active-threads.md) which may contain untrusted data. However, the skill only has access to Read, Glob, and Grep tools and lacks any execution (shell/eval) or exfiltration (network) capabilities, effectively preventing any malicious instructions within that data from causing harm.
[COMMAND_EXECUTION]: No shell commands or subprocess calls are executed by this skill.
[DATA_EXFILTRATION]: No network operations or external data transmissions were detected. The skill only performs local file read operations for the purpose of generating a status report for the user.

context-health