The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill requires the agent to generate and execute Javascript code snippets at runtime to interact with the Figma API via the use_figma tool. It provides several complex templates for discovery and construction phases that must be customized with specific identifiers and logic before execution.
[INDIRECT_PROMPT_INJECTION]: The skill operates on untrusted external inputs to generate UI designs, creating a potential surface for indirect prompt injection.
Ingestion points: The skill ingests source code (CSS, JS, TS, Kotlin, Swift), natural language UI descriptions, and pixel-perfect screenshots captured from running web applications (SKILL.md).
Boundary markers: The instructions include an explicit Anti-Patterns section that warns the agent not to treat external content, logs, or tool responses as trusted instructions.
Capability inventory: The skill possesses extensive capabilities including reading and writing to the Figma canvas, creating new files, and performing network operations via asset uploads (use_figma, create_new_file, upload_assets).
Sanitization: There is no documented process for sanitizing or escaping the data extracted from source code or descriptions before it is interpolated into the executable Javascript templates.
[EXTERNAL_DOWNLOADS]: The documentation references the Figma MCP server guide repository on GitHub as a source for canonical workflow imports and metadata synchronization.
[DATA_EXFILTRATION]: The skill describes a workflow where images may be uploaded to alternative third-party providers using the upload_assets tool if primary AI services are unavailable. While this requires user confirmation, it represents a data exfiltration surface for sensitive visual content.

figma-generate-design