The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's core workflow involves reading technical evidence from the project environment and using it to generate documentation, which creates a potential surface for indirect prompt injection.\n
Ingestion points: According to Step 3 in SKILL.md, the agent is instructed to pull evidence such as current behavior, architecture notes, and operational signals using Read and Grep tools.\n
Boundary markers: The instructions do not define clear boundaries or specific prompts to ensure the agent ignores instructions that may be embedded in the technical data it processes.\n
Capability inventory: The skill is configured with Write and Edit tools in SKILL.md, enabling the agent to commit content influenced by ingested data to the file system.\n
Sanitization: There is no mention of filtering or sanitizing the ingested evidence before it is used in the document generation process.\n- [EXTERNAL_DOWNLOADS]: The skill contains references to several well-known technical documentation standards and communities.\n
The SKILL.md file links to the official websites for Di1taxis, Write the Docs, Architectural Decision Records (ADR), and Keep a Changelog. These references are appropriate for the skill's purpose and originate from established domains.

technical-writing