The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through its data ingestion surfaces.
Ingestion points: The skill reads external design data, including node names, documentation, and component properties, from Figma files via the designs.read permission.
Boundary markers: The skill includes instructions to treat generated implementation code as a visual specification rather than verbatim code. Additionally, the manifest.yaml requires explicit user approval ('ask') for all write operations.
Capability inventory: The skill possesses significant capabilities, including file creation (files.create), design editing (designs.edit), diagram generation (designs.generate), and script execution (use_figma).
Sanitization: The skill relies on user review and visual comparison of output rather than automated sanitization of ingested design text.
[DYNAMIC_EXECUTION]: The use_figma tool enables the agent to generate and execute scripts within the Figma canvas. The skill provides structured rules for script composition to manage this execution surface, which is used for native canvas inspection and mutation.
[COMMAND_EXECUTION]: The skill executes tasks by calling a local figma CLI binary, using commands such as figma call-tool to interact with Figma's infrastructure.
[EXTERNAL_DOWNLOADS]: The skill fetches additional provider-specific guidance and instructions from Figma's official MCP server using skill:// URIs hosted at mcp.figma.com.

figma