image-to-motion

Pass

Audited by Gen Agent Trust Hub on Aug 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local utility scripts for configuration and environment verification (./scripts/check-novoads-env.sh, ./scripts/setup.sh) and performs network operations using curl to interact with the Novoads API (SKILL.md, evals.md).
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied images and transcribes text from them into a structured prompt for the motion model, which introduces a potential surface for indirect prompt injection.
  • Ingestion points: User-provided reference images (SKILL.md).
  • Boundary markers: The prompt template uses a TEXT: header for transcribed content but does not include explicit instructions to the model to ignore commands within that text (SKILL.md).
  • Capability inventory: The agent can execute shell scripts, perform network requests via curl, and write data to the local outputs/ directory (SKILL.md, evals.md).
  • Sanitization: No sanitization is performed on the transcribed text, which is copied verbatim into the prompt (SKILL.md).
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 14, 2026, 12:34 AM
Security Audit — agent-trust-hub — image-to-motion