browser-act

Pass

Audited by Gen Agent Trust Hub on Aug 5, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Installs the browser-act-cli package from PyPI. The documentation notes that the package is distributed as compiled modules, which limits the ability to perform source-level security audits.
  • [COMMAND_EXECUTION]: Executes the browser-act CLI tool to perform browser automation tasks, including navigation and form submissions.
  • [DATA_EXFILTRATION]: Discloses that the CLI transmits telemetry, exception reports, and data required for service features like CAPTCHA solving to the vendor's infrastructure.
  • [PROMPT_INJECTION]: The skill processes untrusted content from websites, creating a surface for indirect prompt injection.
  • Ingestion points: Extracting DOM content and JavaScript-rendered data from websites.
  • Boundary markers: No specific technical delimiters are provided to isolate untrusted web content from agent instructions.
  • Capability inventory: Includes clicks, form submissions, and session management.
  • Sanitization: No mention of sanitizing or escaping data extracted from browsers before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 5, 2026, 01:52 PM
Security Audit — agent-trust-hub — browser-act