browser-automation

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill automates browsing and data extraction from arbitrary web pages, which exposes the agent to instructions embedded in third-party content.
  • Ingestion points: Web page content is captured via screenshots and processed by the AI model during act, assert, and take_screenshot operations.
  • Boundary markers: There are no specific instructions or delimiters mentioned to help the agent distinguish between user-provided tasks and untrusted instructions found on websites.
  • Capability inventory: The skill uses the Bash tool to interact with the file system (screenshots, reports) and control the browser; it can also perform file uploads via the --file-chooser-allowed-dir flag.
  • Sanitization: No sanitization or filtering of website content is described before it is analyzed by the model.
  • [COMMAND_EXECUTION]: The skill requires the Bash tool to execute npx commands and curl for checking local Chrome debugging ports.
  • [EXTERNAL_DOWNLOADS]: The skill fetches the @midscene/web package from the NPM registry at runtime using npx. This is a well-known service and standard for this tool's distribution.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 07:54 AM
Security Audit — agent-trust-hub — browser-automation