The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill generates and writes executable React/TypeScript files (.tsx) into the project's temporary directory (.claude-design/) and the application's route folders (e.g., app/__design_lab/). These generated files are then executed and rendered by the user's development server when navigating to the suggested local URL.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from several untrusted project sources to perform 'Visual Style Inference'.
Ingestion points: Reads project configuration files (package.json, tailwind.config.js, next.config.js), existing component source code, and design memory files (DESIGN_MEMORY.md).
Boundary markers: No explicit delimiters or instructions are used to separate inferred code patterns from the agent's internal logic.
Capability inventory: The agent has the capability to write new components to the filesystem and modify project documentation based on this ingested data.
Sanitization: There is no evidence of sanitization for the content read from existing project files before it is used to generate new code.
[COMMAND_EXECUTION]: The skill instructs the agent to guide the user in running local development commands (e.g., pnpm dev, npm run dev) based on the detected package manager, which facilitates the execution of the dynamically generated design lab components.

design-lab