The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The installation process requires downloading a Figma plugin ZIP file from a personal GitHub repository releases page (https://github.com/vkhanhqui/figma-mcp-go/releases).
[REMOTE_CODE_EXECUTION]: The skill instructs users to install and run the server using npx -y @vkhanhqui/figma-mcp-go@latest, which involves downloading and executing code from the NPM registry provided by an unverified third-party developer.
[COMMAND_EXECUTION]: The skill's operational model relies on executing external commands via npx to start the MCP server, granting the skill execution capabilities on the host system.
[DATA_EXFILTRATION]: Tools like save_screenshots and export_frames_to_pdf allow the agent to write data (images and PDFs) to arbitrary local file system paths via the outputDir and outputPath arguments. This capability could be used to write data to sensitive locations or overwrite existing files.
[INDIRECT_PROMPT_INJECTION]: The skill reads text content from Figma files via tools like get_design_context and scan_text_nodes. If these Figma documents contain malicious instructions, the agent could potentially be manipulated into misusing its file-writing capabilities.
Ingestion points: Figma node properties and text content retrieved from files using get_design_context, scan_text_nodes, and get_node (SKILL.md).
Boundary markers: None. The skill does not implement delimiters or specific warnings to ignore instructions found within the design data.
Capability inventory: The skill has the ability to write to the local file system (save_screenshots, export_frames_to_pdf) and modify live Figma designs.
Sanitization: There is no evidence of input validation or sanitization for the text data retrieved from the design files before it is processed by the agent.

figma-mcp-go