The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs repository maintenance tasks including closing GitHub issues and deleting files. These operations are performed using the gh CLI and filesystem commands. A safety mechanism is in place requiring the user to confirm each individual cleanup action before execution, preventing unauthorized or accidental data loss.
[EXTERNAL_DOWNLOADS]: The skill and its system overview reference documentation hosted on the author's GitHub repository. These references are used for architectural guidance and do not involve the execution of remote scripts or unverified binary downloads.
[PROMPT_INJECTION]: The skill processes user descriptions to diagnose issues, creating an indirect prompt injection surface. Risk is mitigated by the following: 1. Ingestion points: User input is captured in Step 1 of SKILL.md to describe changes. 2. Boundary markers: The skill uses a 'One question per turn' rule and explicit confirmation steps. 3. Capability inventory: The skill uses the gh CLI for issue management and file system access for deletion. 4. Sanitization: Security relies on the human-in-the-loop (HITL) confirmation gate before any destructive action.

correct-course