youtube-to-heygen-longform

Pass

Audited by Gen Agent Trust Hub on Apr 8, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No malicious patterns or security vulnerabilities were detected. The skill correctly uses environment variables to manage sensitive API keys and database credentials, and interacts exclusively with well-known, reputable services (Google Gemini and HeyGen).
  • [PROMPT_INJECTION]: The skill processes untrusted external data (YouTube video content), which presents a surface for indirect prompt injection. This is documented as a vulnerability surface given the skill's capability to send data to external APIs and databases.
  • Ingestion points: YouTube URL content processed by Gemini API in scripts/youtube_to_heygen_longform.py.
  • Boundary markers: Absent; the prompt structure does not use explicit delimiters to isolate the untrusted video content from the system instructions.
  • Capability inventory: External API calls to HeyGen via requests.post and PostgreSQL database writes via psycopg2.
  • Sanitization: Simple whitespace and newline removal is performed on the output script; no semantic content validation is implemented.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 8, 2026, 02:44 PM
Security Audit — agent-trust-hub — youtube-to-heygen-longform