agent-browser
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches workflow instructions and specialized command references (such as those for Slack, Electron, and AWS) from the official Vercel Labs tool repository using the
agent-browser skills getcommand. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to interact with and process data from arbitrary websites and desktop applications, which creates a surface where untrusted content could potentially influence the agent.
- Ingestion points: Web pages and Electron application interfaces accessed via the Chrome DevTools Protocol.
- Boundary markers: The provided stub does not define explicit delimiters for ingested content, relying on the tool's internal snapshotting mechanism.
- Capability inventory: Includes browser automation, session management, video recording, and authentication vault interactions via the
agent-browserCLI. - Sanitization: The underlying tool processes raw page content into accessibility-tree snapshots to simplify and sanitize the data for agent consumption.
Audit Metadata