meshy
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The
depth_local.pyscript performs runtime downloads of binary ONNX model files from HuggingFace and GitHub. These models are essential for the local depth estimation and mesh generation tasks provided by the skill. - [INDIRECT_PROMPT_INJECTION]: The
meshy_batch.pyscript identifies a vulnerability surface for indirect prompt injection when processing batch model generations. - Ingestion points: Manifest JSON files are parsed in
meshy_batch.pyto extract model IDs and prompts for batch processing. - Boundary markers: Prompts are extracted and processed without the use of delimiters or boundary validation instructions.
- Capability inventory: The skill can perform network operations via API calls to Meshy, fal.ai, and other providers, and write to the local file system (e.g., GLB and PLY output).
- Sanitization: Prompt strings from the manifest are passed directly to the provider clients without sanitization or instruction-filtering layers.
Audit Metadata