The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill generates and executes JavaScript code at runtime within Figma by injecting data parsed from external Markdown files into hardcoded templates (Steps 8, 9, and 10). This dynamic assembly is the core mechanism for interacting with the Figma API.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from user-supplied component Markdown files. If these files contain malicious instructions in the '## Color' section or 'render-meta' block, they could influence the agent's actions during the parsing and rendering workflow. Evidence: Ingestion point is the user-provided .md path in Step 0; Capability inventory includes figma_execute and use_figma; Boundary markers are limited to specific HTML comments for metadata; Sanitization is not explicitly defined for the Markdown content.\n- [PROMPT_INJECTION]: The skill uses instruction-override patterns (e.g., 'Never edit this SKILL.md', 'Never call AskQuestion') to establish functional boundaries and prevent the agent from modifying its own code or source documents. While defensive, these patterns match the structure of prompt injection attempts.

create-color