ge-demo-video

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONPRIVILEGE_ESCALATION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes system-level commands to manage the demo generation lifecycle, including Python scripts for font provisioning, browser automation, and video rendering. It also utilizes CLI tools like gcloud and gdrive for cloud storage operations.\n- [EXTERNAL_DOWNLOADS]: Automatically identifies and fetches missing typography assets, such as Google Noto fonts, from official sources to ensure localized demo content renders correctly without missing characters.\n- [INDIRECT_PROMPT_INJECTION]: The skill captures live interaction data from Gemini Enterprise chat sessions via the Chromium DevTools Protocol (CDP), creating an inherent surface for indirect prompt injection.\n
  • Ingestion points: Direct ingestion of chat interface DOM elements (div.ProseMirror) from the live browser session and configuration data from .env files.\n
  • Boundary markers: The instructions do not specify the use of delimiters or specific safety instructions to isolate chat content from the rendering and TTS pipelines.\n
  • Capability inventory: The agent has capabilities to write to the local file system, execute shell commands, and upload data to external cloud storage (Google Drive and GCS).\n
  • Sanitization: There is no explicit description of sanitization or filtering applied to the chat session content before it is processed by the text-to-speech engine or overlayed on video.\n- [PRIVILEGE_ESCALATION]: To facilitate environment setup and video delivery, the skill guides the user to perform privileged actions, such as installing system-level fonts via apt-get and expanding Google Cloud credentials to include Drive access. These actions are managed through explicit interactive approval gates.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 09:18 PM
Security Audit — agent-trust-hub — ge-demo-video