The Agent Skills Directory

[SAFE]: The skill implements secure file path validation in lib/browser-utils.sh using realpath and explicit checks for path traversal patterns, ensuring it only accesses intended files.\n- [SAFE]: Generated HTML files include a Content Security Policy (CSP) and security headers like X-Content-Type-Options: nosniff and X-Frame-Options: DENY, which restrict the browser's execution environment.\n- [SAFE]: Data is handled safely through Base64 encoding in the processing pipeline and rigorous HTML escaping (formatCell) during the rendering phase in templates/scripts/csv-renderer.js.\n- [EXTERNAL_DOWNLOADS]: Fetches typography resources from Google Fonts, a well-known and trusted service. This is documented for transparency but does not pose a security risk.\n- [PROMPT_INJECTION]: The skill's ingestion of external CSV data creates an indirect prompt injection surface. The risk is mitigated by strong boundary markers and sanitization.\n
Ingestion points: CSV content from files or standard input processed in run.sh.\n
Boundary markers: Input content is Base64 encoded before being embedded as a constant string in the output HTML.\n
Capability inventory: The skill uses standard tools for file reading, writing to a dedicated preview directory, and opening the browser.\n
Sanitization: Shell scripts escape filenames for HTML attributes, and the browser-side renderer escapes cell data before insertion into the DOM.

preview-csv