The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates significant repository-wide modifications, including the deletion of files, documentation, and scripts. This is the primary purpose of the skill but represents a high-impact operation on the file system.
[PROMPT_INJECTION]: The skill processes untrusted content from the target repository, creating a surface for indirect prompt injection. Malicious instructions within repository files could attempt to influence the agent's cleanup decisions. 1. Ingestion points: The agent reads architecture docs, plans, issues, and scripts from the repository (SKILL.md Procedure Step 1 & 2). 2. Boundary markers: The skill includes defensive instructions to verify all claims against current code and not to trust metadata or filenames. 3. Capability inventory: The agent has permissions to delete and modify files. 4. Sanitization: No explicit sanitization of repository content is mentioned.

repo-refresh