The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes structured design data from Figma to generate code, which introduces a surface where malicious node metadata or variables could attempt to influence the agent's output behavior.
Ingestion points: Data enters the agent's context through tools like get_design_context, get_metadata, and get_figjam described in references/figma-tools-and-prompts.md.
Boundary markers: The instructions do not define specific delimiters or "ignore instructions" wrappers for the external design data being processed.
Capability inventory: The agent is authorized to generate and modify application code (React + Tailwind) based on the design context.
Sanitization: There are no explicit steps provided for the agent to sanitize or validate the external design metadata before implementation.
[PERSISTENCE]: The configuration reference in references/figma-mcp-config.md suggests persisting the FIGMA_OAUTH_TOKEN environment variable by adding it to shell profiles like ~/.zshrc or ~/.bashrc. This is a standard manual setup procedure for environment persistence in a development environment.

figma