3d-generation-harness
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it is designed to ingest and process untrusted data from external sources.\n
- Ingestion points: Phase 1 (Research Dossier) in SKILL.md and references/dossier-template.md instruct the agent to gather multi-view references and dimension data from external sources (e.g., image searches).\n
- Boundary markers: The process uses structured templates like references/dossier-template.md and a Gate R checklist to delimit and validate data entry.\n
- Capability inventory: The framework coordinates the use of external tools and skills, including blender-assembly, blender-visual-review, and ai-3d-to-dayz for mesh generation, measurement, and auditing.\n
- Sanitization: The harness implements an adversarial review process in Phase 4 (Roast) using references/roast-rubric.md, which serves as a validation step to detect and fix anomalies before final delivery.\n- [NO_CODE]: The skill consists of markdown instructions and templates for a manual workflow. It does not include any executable scripts, binaries, or automated installation commands.
Audit Metadata