agent-browser

Pass

Audited by Gen Agent Trust Hub on May 5, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill describes an agent that visits and processes content from arbitrary web pages, creating a vulnerability to indirect prompt injection.
  • Ingestion points: The agent ingests data from external URLs provided during the intake phase or encountered during navigation (PHASE 1, SKILL.md).
  • Boundary markers: Absent. There are no instructions or delimiters provided to help the agent distinguish between its own system instructions and potentially malicious instructions found on a web page.
  • Capability inventory: The agent can execute JavaScript using eval, take screenshots, and interact with web elements (SKILL.md, Command Reference).
  • Sanitization: No sanitization or filtering of the web content is mentioned before the agent processes it.
  • [COMMAND_EXECUTION]: The instructions explicitly direct the agent to use eval for executing JavaScript code within the browser context.
  • Evidence: The rules section states: NEVER use javascript "..." — JS execution is eval "JS code".
  • This capability provides a high level of control over the browser environment, which could be abused if the agent is influenced by malicious third-party content.
Audit Metadata
Risk Level
SAFE
Analyzed
May 5, 2026, 07:48 AM
Security Audit — agent-trust-hub — agent-browser