The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes command-line tools such as curl and @playwright/cli to perform its tasks.
Evidence: Instructions specify using curl to download screenshot assets and Playwright for browser automation and accessibility snapshots.
[EXTERNAL_DOWNLOADS]: The skill downloads image assets from Figma's infrastructure.
Evidence: Uses curl to fetch temporary screenshot URLs generated by the Figma MCP server (get_screenshot) for local comparison and reporting.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it ingests untrusted text data from Figma design nodes and live website content.
Ingestion points: Figma text nodes (via findAll), Figma metadata, and website accessibility snapshots.
Boundary markers: No specific delimiters are defined for the comparison of text labels and option lists.
Capability inventory: The agent can write to the local file system (runs/ directory), access the network via curl and Playwright, and potentially interface with project management tools like Jira.
Sanitization: There is no explicit instruction to sanitize or escape the text content extracted from external sources before processing or reporting.

figma-compare