video-content-analyzer

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources and is vulnerable to indirect prompt injection attacks.
  • Ingestion points: The scripts/analyze_videos.py script reads video metadata, including titles and captions, from a JSON file provided via the --input argument. This data typically originates from social media scrapers which process attacker-controlled content.
  • Boundary markers: The VIDEO_ANALYSIS_PROMPT interpolates the {caption} variable directly into the instructions without using delimiters (like XML tags or triple quotes) or providing instructions to the AI to ignore potential commands embedded in the caption text.
  • Capability inventory: The skill possesses capabilities that could be abused if an injection is successful, including making arbitrary network requests via the requests library and writing analysis results to the local filesystem using the --output argument.
  • Sanitization: While the script truncates captions to 1000 characters, it performs no escaping, filtering, or validation of the content to ensure it does not contain malicious instructions aimed at overriding the AI's behavior.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 08:18 PM
Security Audit — agent-trust-hub — video-content-analyzer