The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data from Figma files, including text layers and Dev Mode annotations. Instructions in grounding.md explicitly tell the agent to treat these annotations as "authoritative ground truth," creating a vulnerability if a malicious Figma file is provided. (Ingestion points: get_design_context and get_motion_context fetch external node data; Boundary markers: None present to distinguish data from instructions; Capability inventory: Ability to write files and execute shell commands; Sanitization: No sanitization of Figma strings before code generation).
[COMMAND_EXECUTION]: The verification workflow in verify.md directs the agent to execute shell commands such as pnpm build, next build, and pnpm preview to run the generated code in the local environment.
[DYNAMIC_EXECUTION]: The skill generates source code based on external design data and subsequently executes that code via local build tools to perform visual verification against the design source.

figma-codegen