The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill configuration directs the agent to run the MCP server using npx -y @vkhanhqui/figma-mcp-go@latest, which downloads and executes external code from the npm registry at runtime.
[EXTERNAL_DOWNLOADS]: The instructions require users to download a plugin.zip file from an external GitHub repository (github.com/vkhanhqui/figma-mcp-go/releases) and install it locally to enable the Figma plugin bridge.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection by reading untrusted data from Figma files.
Ingestion points: Tools such as get_document, get_selection, and scan_text_nodes pull design content and text metadata into the agent context.
Boundary markers: The skill does not define specific delimiters or instructions to prevent the agent from following directives embedded in the design content.
Capability inventory: The skill has full write access to the Figma file environment and can write files (PDFs, images, tokens) to the host file system.
Sanitization: Content retrieved from Figma is processed without documented validation, filtering, or sanitization to remove potential malicious instructions.
[COMMAND_EXECUTION]: Tools including save_screenshots and export_frames_to_pdf allow the agent to write files to arbitrary local paths provided as parameters.

figma-mcp-go-design-automation