playwright-testing
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to automate browsers and read data from external web pages, which are untrusted sources. Ingestion points: Data enters the agent's context from the browser DOM, console logs, and JavaScript state via Playwright MCP tools. Boundary markers: The skill recommends implementing a stable 'window.TEST' seam to isolate test state and provide a controlled interface. Capability inventory: The agent has the ability to navigate URLs, execute JavaScript in the browser, capture screenshots, and run the provided Python comparison script. Sanitization: The instructions emphasize using structured state snapshots and deterministic readiness signals rather than parsing raw HTML.
- [EXTERNAL_DOWNLOADS]: The 'scripts/imgdiff.py' utility references the 'pillow' library for image processing. This is a well-known, trusted, and standard Python package for visual tasks.
- [DYNAMIC_EXECUTION]: The skill provides a local Python script for image comparison, which is executed as part of the visual regression workflow. This is a standard functional component for the skill's purpose.
- [COMMAND_EXECUTION]: The skill utilizes shell commands to execute the image comparison script and manage test runs.
Audit Metadata