The Agent Skills Directory

Indirect Prompt Injection Surface: The skill processes external data from the local codebase (e.g., source files, styling definitions) to guide design creation. While this is the primary function of the skill, it represents a surface where data could influence the agent's synthesis logic.
Ingestion points: Local project files like *.figma.tsx, *.figma.ts, CSS files, and user-provided UI descriptions (SKILL.md, Step 2a-i).
Boundary markers: The instructions mandate visual validation and font assertion scripts (references/discover-product-font.md) to ensure output fidelity.
Capability inventory: The skill uses use_figma for script execution, search_design_system for library lookups, and generate_figma_design for web captures.
Sanitization: The agent acts as the translator, interpreting code patterns into Figma API calls, which naturally limits the execution context of the ingested data.
Dynamic Script Generation: The skill generates JavaScript code at runtime to interact with the Figma API. This process uses predefined patterns and discovered component keys to automate design tasks, which is a standard implementation for Figma-agent integrations.
Local Resource Discovery: The workflow involves scanning the codebase for design system links (e.g., Code Connect files) and styling tokens. This access is targeted and limited to the metadata required to link code components to Figma library items.

figma-generate-design