shot-by-shot-analysis

Pass

Audited by Gen Agent Trust Hub on Sep 11, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external media which could potentially contain hidden instructions intended to influence the agent's behavior.
  • Ingestion points: The workflow involves inventorying and inspecting multiple input videos from external paths or URLs (SKILL.md).
  • Boundary markers: The skill mandates the use of a strict Beat Evidence Contract and a specific Markdown template (references/report-template.md) to ensure isolated analysis. It also includes an Imitation Boundary section in SKILL.md to prevent the adoption of specific identities or brand marks.
  • Capability inventory: The skill invokes the media-analysis tool to perform video understanding tasks (SKILL.md).
  • Sanitization: While the skill requires precision and uncertainty markers for speech and text, it does not specify methods for sanitizing or escaping content extracted from videos that might contain adversarial prompts.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 11, 2026, 01:16 AM
Security Audit — agent-trust-hub — shot-by-shot-analysis