The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to clone a repository from an untrusted external source: https://github.com/tercumantanumut/sunnysideFigma-Context-MCP.
[REMOTE_CODE_EXECUTION]: The installation steps include running npm install and npm start on the cloned repository. This represents a high-severity risk of arbitrary code execution from an unverified third party.
[COMMAND_EXECUTION]: The skill requires several shell commands to be executed for setup and operation, including git clone, npm install, npm run build, and npm start.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from Figma designs, which are external and potentially attacker-controlled, posing a risk of malicious instructions being processed by the agent.
Ingestion points: Tools such as get_JSON and get_figma_data fetch content from the Figma API and a local plugin bridge.
Boundary markers: No explicit delimiters or warnings for the agent to ignore instructions within the design data are mentioned.
Capability inventory: The skill includes file-writing capabilities (download_figma_images) and generates code/styles from external inputs.
Sanitization: The documentation does not indicate that text or metadata from Figma layers are sanitized before being presented to the agent.

sunnyside-figma-context-mcp