The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references several external Node.js packages and official documentation from trusted and well-known sources.
Evidence: References to better-all, lru-cache, swr, zod, lucide-react, and @mui/material in AGENTS.md and individual rule files.
Source Status: Resources are linked to trusted organizations (Vercel) and well-known open-source maintainers.
[COMMAND_EXECUTION]: The skill mentions the use of a CLI tool for asset optimization.
Evidence: The rule rules/rendering-svg-precision.md suggests using npx svgo to reduce SVG precision.
Context: This is a standard and benign development practice for performance optimization.
[INDIRECT_PROMPT_INJECTION]: The skill is intended to be used for reviewing and refactoring external codebases, which represents a potential attack surface.
Ingestion points: External React and Next.js code files (components, pages, API routes) processed by the agent during tasks such as Writing new React components or Reviewing code for performance issues as described in SKILL.md.
Boundary markers: The skill instructions do not provide specific delimiters or warnings for the agent to ignore instructions embedded within the code being analyzed.
Capability inventory: Across its scripts, the skill grants the agent the ability to generate code performing network operations (fetch, SWR), modify files during refactoring, and execute CLI tools (svgo).
Sanitization: No explicit sanitization or validation of the processed code is mentioned in the skill guidelines.

vercel-react-best-practices