The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is configured with access to Bash for commands beginning with npx, ls, and cat. This allows the agent to execute Node.js packages and explore the local file system to gather context for design reviews.
[EXTERNAL_DOWNLOADS]: The skill uses npx to run utilities, which typically involves fetching packages from the npm registry. Additionally, the 'Phase 1: Visual Inspection' step involves navigating to external URLs via Playwright to capture screenshots and accessibility data.
[PROMPT_INJECTION]: The skill identifies an indirect prompt injection surface by processing content from external URLs. An attacker could potentially place malicious instructions on a webpage that the agent is asked to review, which might influence the agent's behavior during the session.

designer