omh-model-compression
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches documentation and code examples from official repositories and documentation sites including GitHub (NVIDIA, ggml-org, vllm-project), HuggingFace, and PyTorch.
- [COMMAND_EXECUTION]: Provides instructions for executing standard model compression and evaluation tools such as llama-quantize, llama-imatrix, and lm_eval.
- [INDIRECT_PROMPT_INJECTION]: Ingests user-provided task examples and calibration text for model evaluation and quantization.
- Ingestion points: User-provided datasets and examples referenced in SKILL.md and references/compression-method.md.
- Boundary markers: Not explicitly specified for the content of the ingested data.
- Capability inventory: Utilizes ML libraries (llmcompressor, llama.cpp, transformers) to process data and modify model weights.
- Sanitization: No explicit sanitization of dataset content is described.
Audit Metadata