The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads the @playwright/cli package from the NPM registry to perform UI state audits.
[REMOTE_CODE_EXECUTION]: Employs npx to fetch and run the Playwright testing utility at runtime.
[COMMAND_EXECUTION]: Runs shell commands to interact with the browser and save audit evidence to a local directory.
[PROMPT_INJECTION]: The skill ingests and evaluates content from external websites, providing a surface for indirect prompt injection.
Ingestion points: External application content retrieved via the $APP variable during browser navigation in Phase 2.
Boundary markers: There are no instructions to differentiate or ignore embedded content during the audit process.
Capability inventory: The agent can execute shell commands via npx and write files to the .playwright-mcp/ directory.
Sanitization: Content observed from the application is processed without defined filtering or sanitization steps.

audit-ui-states