The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a local Python script located at {WORKSPACE}/skills/therapy-mode/therapy-notes.py for managing therapy notes through various subcommands (new, add, insight, etc.). This execution is legitimate and scoped to the user's workspace environment.\n- [PROMPT_INJECTION]: The skill contains 'hardened' instructions designed to override potential adversarial prompts that attempt to skip crisis assessments. While these are meta-instructions used for safety enforcement, they utilize override patterns to govern agent behavior.\n- [PROMPT_INJECTION]: The therapy-notes.py add <text> command represents a surface for indirect prompt injection, as the agent is instructed to pass user-provided session dialogue into a command-line argument. This is documented as a vulnerability surface where user-supplied content enters the execution flow.\n- [SAFE]: The skill maintains sensitive therapy data within the local workspace directory {WORKSPACE}/therapy-notes/. No evidence of unauthorized data exfiltration or credential exposure was detected.\n- [SAFE]: External references to faberlens.ai are official documentation links providing transparency on the skill's safety evaluation and do not represent a security risk.

therapy-mode-hardened