The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted design data from Figma artboards, variables, and metadata to generate application source code. While this creates an attack surface where a malicious Figma file could attempt to influence the agent's logic, the risk is inherent to the skill's primary function.
Ingestion points: figma_get_metadata, figma_get_design_context, and figma_get_variable_defs (Phase 1).
Boundary markers: The instructions emphasize strict adherence to component documentation but do not provide explicit delimiters or warnings to ignore instructions within the design data.
Capability inventory: The agent can execute shell commands (npx, curl, npm), modify the local project filesystem, and run JavaScript in a browser context via Playwright.
Sanitization: Figma-derived data is used directly in code generation without specific sanitization or filtering steps mentioned.
[DATA_EXPOSURE]: The skill manages a FIGMA_ACCESS_TOKEN and provides clear instructions for the user to store this sensitive credential in environment variables or IDE-specific configuration files like .vscode/mcp.json. The agent accesses these during Phase 0 and Phase 1h for environment verification and asset extraction.
[EXTERNAL_DOWNLOADS]: Automated asset extraction in Phase 1h utilizes curl to fetch images and icons from the official Figma API (api.figma.com). These downloads target well-known infrastructure and are essential for maintaining design fidelity.
[COMMAND_EXECUTION]: Phase 5 uses Playwright to conduct visual audits and layout measurements of the running Angular application. This involves executing predefined JavaScript diagnostic functions within the browser via playwright_browser_evaluate to verify CSS properties and DOM structure.

igniteui-angular-figma-to-app