jianying-draft-motion

Pass

Audited by Gen Agent Trust Hub on Oct 26, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external JSON data structures, specifically interaction events and smart-zoom plans, which serves as a potential attack surface for indirect prompt injection.
  • Ingestion points: The skill reads structured data from interaction-events-v1.json and smart-zoom-plan-v1.json to generate video editing plans.
  • Boundary markers: Instructions mandate that all media facts, durations, and project settings must be verified against tool-derived output rather than relying on the input data alone.
  • Capability inventory: Capabilities are restricted to a specific Rust CLI tool (jianying), and the skill is explicitly prohibited from generating Python scripts or calling external headless environments.
  • Sanitization: Input data is validated against strict JSON schemas, and coordinates are normalized within the capture area. Time-based inputs are subjected to frame quantization to ensure they align with the project grid.
  • [SAFE]: The skill implements strong security best practices, such as requiring isolated output paths to prevent accidental overwriting of source files and explicitly forbidding the capture of sensitive keyboard content.
  • [SAFE]: Data handling is restricted to the local environment and declared provider boundaries, with clear instructions to avoid collecting or logging unrelated credentials.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 26, 2026, 07:55 PM