web-visual-verification
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions and provided code skeletons utilize the well-known
playwrightNode.js package for browser automation tasks. - [INDIRECT_PROMPT_INJECTION]: The skill templates are designed to ingest and process content from external websites, which presents a surface for indirect prompt injection if the targeted site contains adversarial instructions.
- Ingestion points: Data enters the execution context through
page.goto(URL)andpage.evaluate(), specifically when extracting page text in scripts likescripts/check-text-absence.mjs. - Boundary markers: The templates do not currently include explicit delimiters or instructions for the agent to ignore potentially malicious content embedded in the target web pages.
- Capability inventory: The generated scripts have the capability to perform network requests via Playwright, create directories, and write screenshots to the local filesystem using
node:fs/promises. - Sanitization: The provided logic does not include sanitization or validation of the retrieved web content before it is processed or logged.
Audit Metadata