The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized data access were detected. The skill's behavior aligns strictly with its stated purpose of implementing visual testing.
[EXTERNAL_DOWNLOADS]: The skill references standard resources such as Microsoft Playwright Docker images and Chromatic GitHub Actions. These originate from well-known services and are used according to standard practices.
[INDIRECT_PROMPT_INJECTION]: Evaluation of the untrusted data ingestion surface:
Ingestion points: Page content is loaded via page.goto() in SKILL.md and recipes.md snippets.
Boundary markers: Not used, as the content is processed for visual comparison rather than semantic instruction execution.
Capability inventory: Shell command execution via npx playwright and screenshot file creation.
Sanitization: The skill recommends masking dynamic content and freezing the system clock, which provides determinism and reduces the risk of processing unexpected visual data.

visual-testing