The Agent Skills Directory

[SAFE]: The skill provides instructions for a phased documentation workflow (Discover, Research, Outline, Draft, Verify). All identified actions, such as reading repository files and performing web searches for technical facts, are standard and necessary for its documented purpose.
[PROMPT_INJECTION]: The skill ingests untrusted data from external sources (web search) and internal repository files. While this represents a surface for indirect prompt injection, the agent is instructed to focus on extracting technical facts and generating markdown prose, which significantly limits the impact of potential injections compared to skills that execute code.

documentation-writer