genimg-gemini-web

Pass

Audited by Gen Agent Trust Hub on Mar 5, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: Spawns a local Chrome or Chromium browser to facilitate authentication with Gemini Web. This is a core functional requirement used solely to synchronize session cookies.\n- [EXTERNAL_DOWNLOADS]: Fetches generated images, videos, and configuration data from official Google domains (google.com, googleusercontent.com).\n- [PROMPT_INJECTION]: The skill ingests data from external sources and interpolates it into the LLM prompt, creating a surface for indirect prompt injection.\n
  • Ingestion points: Untrusted content is read from command-line arguments, stdin, and local files via --promptfiles and --reference flags in scripts/main.ts and scripts/executor.ts.\n
  • Boundary markers: No delimiters or isolation instructions are present to distinguish user-provided content from system instructions.\n
  • Capability inventory: The skill can execute local browser processes, perform authenticated network requests to Google services, and write cookies, sessions, and media files to the local file system.\n
  • Sanitization: External content is included in the prompt without escaping or validation.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 5, 2026, 07:16 AM