ingest-video

Pass

Audited by Gen Agent Trust Hub on Jul 21, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute a simple sleep 30 command via Bash to manage polling intervals. This is a controlled execution of a benign system command.
  • [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by processing and summarizing video transcripts fetched from an external source. Ingestion points: Processes content from user-provided URLs and retrieved video transcripts. Boundary markers: None explicitly mentioned for the transcript content. Capability inventory: Uses MCP tools (upload_video, get_upload_status, get_project, get_transcript) and executes sleep 30 in bash. Sanitization: Not mentioned.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 21, 2026, 01:14 PM
Security Audit — agent-trust-hub — ingest-video