video-artifact-intake

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill invokes ffprobe to extract video metadata such as duration, dimensions, and codecs. This is a legitimate use of the tool for video analysis.
  • [COMMAND_EXECUTION]: The skill executes local Node.js scripts (scripts/harness/video-intake.ts and ./.video-evaluator/agent/run-tool.mjs) to perform its primary logic. These scripts are part of the skill's own package or the target repository.
  • [PROMPT_INJECTION]: Indirect Prompt Injection Surface: The skill reads and processes external JSON artifacts produced by the repository which could theoretically contain instructions to influence agent behavior.
  • Ingestion points: Local files such as quality.json, verification.json, latest.json, and timeline.evidence.json as specified in SKILL.md.
  • Boundary markers: None identified in the documentation.
  • Capability inventory: Local file system reads and metadata extraction via subprocesses (ffprobe).
  • Sanitization: No explicit sanitization or validation of the content within these JSON reports is described.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 02:03 AM
Security Audit — agent-trust-hub — video-artifact-intake