The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install various @farming-labs/* packages and use npx @farming-labs/docs@latest init to scaffold projects. It also supports fetching OpenAPI specifications from external URLs (e.g., petstore3.swagger.io) for API reference generation. These actions are consistent with the framework's purpose.
[COMMAND_EXECUTION]: The skill describes using CLI tools for project initialization (npx @farming-labs/docs@latest init), documentation review (docs review), and generating static assets (docs agents generate, docs robots generate). These commands are part of the standard development workflow for this framework.
[INDIRECT_PROMPT_INJECTION]: The framework is specifically designed to expose project documentation to AI agents through machine-readable formats like AGENTS.md, llms.txt, and dedicated Markdown routes. This creates a surface where an agent could ingest and potentially follow malicious instructions embedded in the documentation source files.
Ingestion points: Documentation content is ingested from Markdown and MDX files stored in the project's content directory (e.g., docs/ or app/docs/).
Boundary markers: The framework uses <Agent> and <Human> tags to differentiate content, but these act as visibility filters rather than secure boundaries against adversarial content.
Capability inventory: The generated documentation (including API references and changelogs) is intended to be consumed by coding agents to facilitate project understanding and task execution.
Sanitization: There is no mention of sanitization or validation processes to filter out prompt injection attacks within the documentation content itself.

getting-started