The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides detailed instructions for generating JavaScript code snippets intended for execution via the use_figma tool. These scripts are used to automate design tasks such as font loading and node creation. This dynamic execution is a core feature of the integration and is restricted to the Figma API environment.
[EXTERNAL_DOWNLOADS]: The skill references image assets retrieved via platform-managed URLs (e.g., https://…/api/mcp/asset/<uuid>) through the design context tool. These operations are part of the standard asset retrieval workflow for the Model Context Protocol (MCP).
[PROMPT_INJECTION]: The skill processes untrusted external data from Figma files, including node metadata and 'Code Connect' snippets, which can influence the generated SwiftUI code or Figma scripts.
Ingestion points: Figma node data and code snippets are ingested via the get_design_context tool as described in references/design-to-code.md.
Boundary markers: The instructions distinguish between structural references and authoritative sources, providing a conceptual boundary for the AI agent.
Capability inventory: The agent has the capability to write local SwiftUI source files and execute script-based commands in the Figma environment.
Sanitization: The skill employs a mapping-based translation logic that focuses on architectural patterns rather than raw string execution, mitigating the risk of direct command injection from the design data.

figma-swiftui