agent-browser
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill allows the agent to navigate to arbitrary URLs and extract text content, which introduces a surface for indirect prompt injection. Malicious instructions placed on web pages could be ingested by the agent and influence its subsequent actions.
- Ingestion points: Page content extracted using
agent-browser textandagent-browser snapshot. - Boundary markers: The instructions do not specify any delimiters or special handling to differentiate between web content and agent instructions.
- Capability inventory: The skill can navigate to any URL, interact with page elements (click/type), take screenshots, and write content to local files (e.g.,
article.txt). - Sanitization: No sanitization or safety filtering of the extracted web content is mentioned.
- [EXTERNAL_DOWNLOADS]: The skill documentation guides the user to install the
agent-browserCLI tool through npm or Homebrew. These downloads originate from Vercel Labs, which is a well-known and recognized service provider.
Audit Metadata