The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell scripts (scripts/figma-export.sh, scripts/visual-diff.sh) to wrap system commands such as curl, python3, and ImageMagick. While these scripts handle user-provided inputs like node IDs, they employ sanitization (e.g., via sed) and safe parameter passing (e.g., curl --data-urlencode) to mitigate injection risks.
[DYNAMIC_EXECUTION]: scripts/render-capture.mjs enables the execution of arbitrary JavaScript and CSS within the Playwright-controlled browser via the INIT_SCRIPT and NEUTRALIZE_CSS environment variables. This is an intended feature for freezing UI state (e.g., mocking Date.now) but provides a vector for code execution within the browser environment.
[DYNAMIC_EXECUTION]: scripts/render-capture.mjs performs dynamic imports of @playwright/test or playwright relative to the current working directory. This allows the tool to use the project's installed dependencies but depends on the security of the local node_modules environment.
[INDIRECT_PROMPT_INJECTION]: The skill captures screenshots of arbitrary web URLs provided by users. This creates an indirect prompt injection surface, as the agent processes the visual and structural results of these external pages, which could contain malicious content designed to influence the agent's analysis.
[EXTERNAL_DOWNLOADS]: scripts/figma-export.sh downloads reference images from URLs returned by the Figma REST API. These external resources originate from Figma's trusted image hosting infrastructure.
[CREDENTIALS_UNSAFE]: The skill manages Figma API tokens through environment variables (FIGMA_TOKEN, FIGMA_API_KEY) rather than hardcoding them. The implementation includes explicit warnings against passing secrets as command-line arguments.

design-to-code-fidelity