interactive-browser-testing

Pass

Audited by Gen Agent Trust Hub on Sep 8, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external websites, creating a surface for indirect prompt injection attacks.
  • Ingestion points: The skill ingests data from external URLs via browser_navigate, browser_snapshot, browser_take_screenshot, and browser_console_messages (SKILL.md).
  • Boundary markers: There are no explicit delimiters or instructions provided to the agent to disregard instructions embedded within the processed web content or console logs.
  • Capability inventory: The skill possesses powerful capabilities including browser_evaluate (arbitrary JavaScript execution), browser_type, browser_file_upload, and web_search via the mcpc CLI.
  • Sanitization: No sanitization or validation logic is defined for the content retrieved from the browser before it is passed to the AI analysis tool (@minimax) or used to inform subsequent tool calls.
  • [COMMAND_EXECUTION]: The skill uses a command-line interface (mcpc) to interact with browser and AI tools.
  • The browser_evaluate tool (SKILL.md) allows for the execution of arbitrary JavaScript within the context of the navigated website, which is a powerful capability that could be misused if the agent is compromised by malicious page content.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 8, 2026, 10:53 AM
Security Audit — agent-trust-hub — interactive-browser-testing