The Agent Skills Directory

[DATA_EXPOSURE]: The skill includes a specific safety rule to avoid writing absolute filesystem paths in documentation, instructing the agent to use repository-relative paths instead to prevent revealing sensitive local identity details.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from a repository (such as AGENTS.md and codebase content) to generate documentation. While this creates a surface for indirect prompt injection, the skill mitigates this by instructing the agent to verify all claims against multiple sources of truth (the 'evidence ladder') and focuses purely on documentation output.

docs