The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests existing project documentation and repository structure (SKILL.md) to inform its generation process. This creates a surface where malicious instructions in repository files could influence the agent. The skill implements a structured 6-question interactive boundary and maintains a file-writing capability to the project root. While explicit data sanitization is absent, the mandatory human-in-the-loop confirmation before file operations serves as a significant safeguard.\n- [SAFE]: The skill requires user interaction and confirmation before overwriting or modifying files, providing options for merging or backups to prevent accidental data loss.\n- [SAFE]: The skill explicitly instructs contributors to never commit secrets, adhering to repository security best practices.

200-agents-md