The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied OpenAPI specifications and MDX documentation files, which creates an attack surface for indirect prompt injection. \n
Ingestion points: OpenAPI specifications (openapi.yaml), MDX files (fern/**/*.mdx), and YAML configuration files (docs.yml, product.yml). \n
Boundary markers: Relies on the inherent structure of MDX and YAML files; no specific delimiter-based boundary instructions are provided for the agent's context. \n
Capability inventory: Executes fern CLI commands for project initialization and publishing, performs file reads/writes, and can add MCP servers. \n
Sanitization: Instructions guide the agent to use fern check for validation and to verify behaviors against official documentation rather than fabricating data. \n- [COMMAND_EXECUTION]: Uses shell commands to interact with the Fern ecosystem, including initialization (fern init), validation (fern check), and site generation (fern generate). It also uses npm install for CLI setup and grep for searching repository content. These commands are necessary for documentation management. \n- [EXTERNAL_DOWNLOADS]: Fetches CLI tools, configurations, and schemas from the vendor's official domain (buildwithfern.com) and GitHub organization (fern-api). These are trusted resources belonging to the skill's author. \n- [SAFE]: The skill includes explicit instructions that the repository should never hold client secrets or signing keys. It also recommends providing users with production commands to run themselves after checking preview URLs, rather than the agent executing them autonomously without confirmation.

fern-docs