The Agent Skills Directory

[SAFE]: The skill automates design analysis by using Playwright to extract visual styles from user-provided URLs. This process is transparent and focused on design tokens.
[SAFE]: The workflow utilizes the --isolated flag for browser sessions, ensuring that no sensitive user data (like cookies or login sessions) is accessed or leaked during the analysis.
[SAFE]: Installation instructions use standard platform commands to add the Playwright MCP server, a well-known tool for browser automation.
[COMMAND_EXECUTION]: Local shell execution (bash) is used to run grep for a self-audit pass (filtering out generic AI phrases) and python3 is used for JSON validation. These are benign operations intended to ensure output quality.
[SAFE]: The skill writes design maps and tool-specific rules to the local filesystem as intended by its core functionality. These exports are directed to standard configuration paths for tools like Cursor, Windsurf, and Claude Code.
[SAFE]: The multi-step analysis pipeline includes a dedicated 'Observer' agent that serves as a quality gate, ensuring the final output is evidence-based and free of generic descriptions.

taste