The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill establishes rules for agents to refactor and generate React/Next.js codebases.
Ingestion points: The agent is designed to process user-provided source code files (SKILL.md).
Boundary markers: The instructions do not define specific delimiters or warnings to disregard instructions potentially hidden in the target code being optimized.
Capability inventory: The agent uses file-writing and code-generation capabilities to apply these performance rules across the project.
Sanitization: No explicit logic for filtering user code for adversarial prompt patterns is included.
[DYNAMIC_EXECUTION]: The skill includes code templates for performance optimization that utilize dynamic execution patterns.
Evidence: Rule rendering-hydration-no-flicker.md provides a correct implementation example using dangerouslySetInnerHTML to inject a synchronous <script> block into the DOM for theme management. This is a recognized performance pattern in the Next.js ecosystem for preventing visual flicker.
[EXTERNAL_DOWNLOADS]: The instructions recommend several third-party libraries for specialized tasks.
Evidence: References include better-all (parallelizing partially dependent promises), lru-cache (cross-request caching), swr (client-side data fetching and revalidation), and zod (schema validation). All referenced packages are well-known and widely used within the professional web development community.
[COMMAND_EXECUTION]: The skill suggests the use of command-line utilities for asset optimization.
Evidence: Rule rendering-svg-precision.md recommends executing npx svgo to automate the reduction of coordinate precision in SVG files. This is a standard optimization utility.

vercel-react-best-practices