blender-tracking
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is composed strictly of documentation, process checklists, and configuration templates. It does not include any scripts, binaries, or commands that would interact with the network or sensitive local filesystem locations.
- [INDIRECT_PROMPT_INJECTION]: The skill describes a workflow that ingests external data (video clips and marker coordinates), creating a surface for indirect prompt injection. The skill mitigates this through prescriptive validation steps:
- Ingestion points: Loading of media clips and named tracking coordinates from external files (SKILL.md).
- Boundary markers: Instructions require clips to be loaded only from 'approved asset roots' and coordinates to be explicitly normalized (SKILL.md).
- Capability inventory: Usage is scoped to internal Blender operators for camera solving and scene setup (SKILL.md).
- Sanitization: Mandatory result validation requires measuring reprojection error and bundle counts against a specific threshold, preventing the acceptance of malicious or faulty data as a 'solve' (SKILL.md).
Audit Metadata