resolve-rough-cut

Pass

Audited by Gen Agent Trust Hub on Aug 31, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes ffprobe to extract media metadata (frame rates, rotation) and executes a local script scripts/contact_sheet.py via python3 to generate visual summaries of footage. These actions are restricted to the local environment and the DaVinci Resolve API for project management and timeline assembly.
  • [EXTERNAL_DOWNLOADS]: The documentation identifies a dependency on the Pillow Python library for image processing tasks. Installation is suggested via pip, referencing a well-known and trusted package.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes raw video footage which acts as an untrusted data source.
  • Ingestion points: Raw behind-the-scenes and vlog footage files processed from local directories (SKILL.md).
  • Boundary markers: None explicitly defined for the analysis phase; however, the workflow uses structured API calls rather than direct prompt interpolation of data.
  • Capability inventory: Includes shell command execution (ffprobe, python3), file system writes for contact sheets and renders, and project modification through the DaVinci Resolve API.
  • Sanitization: No specific content sanitization is described for the visual frames extracted during media analysis.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 31, 2026, 10:28 PM