The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use standard system commands (open, start, or xdg-open) to display generated artifacts to the user. This is the primary intended function for viewing the standalone HTML canvases. It also involves using npm commands to build and serve the technical reader locally.
[EXTERNAL_DOWNLOADS]: The design canvas template (design.md) incorporates styling and diagramming libraries (Tailwind CSS, Mermaid, and Chart.js) via well-known and trusted content delivery networks (JSDelivr).
[INDIRECT_PROMPT_INJECTION]: The technical reader is designed to ingest and render Markdown content which may be sourced from untrusted external data. The skill proactively mitigates this risk through a defense-in-depth sanitization strategy.
Ingestion points: The scripts/reader/src/report.md file serves as the content source for the React application (imported in src/main.tsx).
Boundary markers: The instructions emphasize distinguishing fictional examples from verified evidence and require specific table structures for task tracking.
Capability inventory: The reader application (scripts/reader/src/App.tsx) is a client-side tool with no backend execution capabilities, although it uses the Clipboard API for utility functions (scripts/reader/src/CopyButton.tsx).
Sanitization: The implementation uses rehype-sanitize for the Markdown body in scripts/reader/src/Markdown.tsx and DOMPurify to cleanse SVG output from Mermaid diagrams in scripts/reader/src/Diagram.tsx, specifically stripping active content like scripts, links, and foreign objects.

canvas