The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes component data from Figma nodes which could potentially contain malicious instructions intended to influence the generated code or the agent's workflow.
Ingestion points: Figma node metadata, component names, and property definitions are retrieved through the get_context_for_code_connect and get_code_connect_suggestions tools.
Boundary markers: The skill does not explicitly use specific delimiters to isolate Figma-provided data within the generated templates.
Capability inventory: The skill can search the local codebase for components, write new .figma.ts files, and prompt the execution of CLI commands like npx figma connect publish.
Sanitization: While the skill uses figma.code tagged templates for formatting the output, it does not explicitly validate or sanitize the raw string values retrieved from Figma properties before interpolation.
[DYNAMIC_EXECUTION]: The core functionality of the skill involves generating executable template files based on external metadata.
Evidence: Instructions in SKILL.md and the references/ directory guide the creation of .figma.ts and .figma.js files that utilize the figma API.
Context: These scripts are generated from design metadata and are intended for processing by the official Figma Code Connect CLI utility.
[EXTERNAL_DOWNLOADS]: The skill references external tools and repositories necessary for the Code Connect workflow.
Evidence: The documentation recommends installing the @figma/code-connect package from the NPM registry and references the official Figma MCP server guide on GitHub.
Trust: These resources originate from the figma organization, which is a recognized and well-known service provider.

figma-code-connect