videodb

Pass

Audited by Gen Agent Trust Hub on Apr 11, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute Python scripts and manage a background WebSocket listener (ws_listener.py). This is required for its core functionality of capturing real-time desktop sessions and processing video streams.
  • [EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the videodb Python SDK and python-dotenv from standard package registries. These are necessary dependencies for interacting with the VideoDB service.
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection as it processes untrusted data such as video transcripts and visual descriptions through LLM integration (coll.generate_text).
  • Ingestion points: External video content is ingested via video.get_transcript_text() and video.index_scenes().
  • Boundary markers: Prompts provided in examples lack explicit delimiters (e.g., XML tags or triple backticks) to wrap the untrusted transcript content.
  • Capability inventory: The skill has access to file system writes (local events log) and network operations (VideoDB API calls).
  • Sanitization: There is no evidence of transcript content sanitization or filtering before it is sent to the LLM for summarization or analysis.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 11, 2026, 09:17 AM