The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use standard Unix utilities such as find, cat, ls, and mkdir to inspect the project structure and create documentation files. These commands are used for their intended purpose within the local development environment and do not involve high-risk operations.
[SAFE]: The skill reads project metadata from package.json and scans for .env.example files to document environment requirements. It does not attempt to access sensitive credential files such as .env, ~/.ssh, or AWS configuration directories.
[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection because it processes content from the project codebase (e.g., package.json fields, directory names, and TypeScript exports) to populate documentation templates.
Ingestion points: Project package.json, .env.example, and file system listings (via find and ls) are read in references/workflows/docs.md to gather context.
Boundary markers: The workflow does not explicitly define delimiters or instructions to the agent to disregard potentially malicious instructions embedded in the project files during the templating process.
Capability inventory: The agent is authorized to create directories and write markdown files within the docs/ directory of the workspace.
Sanitization: The skill does not perform specific validation or sanitization of the ingested metadata before inserting it into the templates. However, given the primary purpose of the skill is documentation and the impact is limited to markdown generation, this is considered a low-risk, functional requirement.

dev-docs