The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a Python-based web server (scripts/viewer.py) in the background to serve documentation locally or on the LAN.
[INDIRECT_PROMPT_INJECTION]: The skill processes and renders external files (Markdown, CSV, JSON). This surface is mitigated by comprehensive HTML entity escaping in assets/app.js and a restrictive Content Security Policy (CSP) defined in scripts/viewer.py.
Ingestion points: The viewer.py script reads file contents via Path.read_text and serves them through the /api/doc endpoint.
Boundary markers: The skill is intended for documentation viewing; rendered content is separated from the application logic via CSP.
Capability inventory: The skill has read access to served files but does not include file-write or external network exfiltration capabilities.
Sanitization: Content is sanitized using a custom escaping function for HTML, Mermaid's strict mode, and by disabling raw HTML rendering in the markdown parser.
[SAFE]: The server includes a built-in security filter (is_secret function in viewer.py) that explicitly blocks access to sensitive file patterns such as .env files, SSH private keys, and cloud provider credentials.
[SAFE]: Access to the served documentation requires a randomly generated 16-character token, which is passed via URL and subsequently stored in an HttpOnly cookie to limit exposure in browser history.

lan-viewer