effective-web-research

Pass

Audited by Gen Agent Trust Hub on Jul 27, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides specific instructions for the agent to execute shell commands using curl to interact with a local CDP proxy (web-access) at http://localhost:3456. This interface allows the agent to send JavaScript for evaluation, take screenshots, and simulate clicks within a browser environment.
  • [PROMPT_INJECTION]: The skill is designed to ingest and process data from the external web, creating a surface for indirect prompt injection. Malicious instructions embedded in fetched web pages could attempt to influence the agent's behavior or leverage the browser access capabilities.
  • Ingestion points: External web pages fetched via WebSearch, WebFetch, curl, and the web-access CDP proxy.
  • Boundary markers: The skill uses a structured reporting format but lacks explicit technical delimiters or directives to the agent to ignore instructions embedded in the external content.
  • Capability inventory: The agent has capabilities to perform web searches, fetch content, and interact with a local browser proxy that carries login state and can execute arbitrary JavaScript in the context of the user's browser.
  • Sanitization: The skill relies on qualitative evaluation by the agent rather than technical sanitization or filtering of the retrieved content.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 27, 2026, 04:59 PM
Security Audit — agent-trust-hub — effective-web-research