The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several command-line tools including playwright-cli and firecrawl to perform web discovery, headless browsing, and data extraction from specified URLs.
[EXTERNAL_DOWNLOADS]: The skill references and downloads legitimate developer tools, including @playwright/cli from Microsoft and the @google/design.md linter. These originate from trusted organizations and well-known registries.
[REMOTE_CODE_EXECUTION]: The skill uses npx to dynamically fetch and execute the @google/design.md linting tool. This is a standard operation for performing quality checks on the generated design files using an official tool from a trusted vendor.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it ingests untrusted data from external websites to generate design documentation. However, the risk is minimal as the ingested content is used solely for documentation synthesis and does not drive sensitive system operations.
Ingestion points: External website content (CSS variables, HTML elements, and text copy) is read via Playwright and Firecrawl throughout Stages 1-4.
Boundary markers: No specific delimiters or instructions are used to isolate scraped content during the Markdown synthesis phase in Stage 5.
Capability inventory: The skill has access to shell execution (playwright-cli, npx, firecrawl), but these capabilities are utilized for the extraction workflow and are not influenced by the processed data.
Sanitization: The skill does not perform explicit sanitization or validation of the scraped website content before including it in the generated DESIGN.md file.

extract-design-md