gptq

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches machine learning model weights and datasets from HuggingFace, a well-known service in the AI community (e.g., models from 'TheBloke' and 'meta-llama', and the 'c4' dataset).
  • [EXTERNAL_DOWNLOADS]: Provides instructions to install standard, community-verified libraries including 'auto-gptq', 'transformers', 'peft', and 'vllm' from official package registries.
  • [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow that ingests untrusted external data (such as the 'c4', 'ShareGPT', or 'the-stack' datasets) to calibrate the quantization process.
  • Ingestion points: Use of 'datasets.load_dataset' within 'SKILL.md' and 'references/calibration.md' to pull remote text samples.
  • Boundary markers: None; the raw text is tokenized directly for mathematical importance calculations.
  • Capability inventory: The ingested data is used solely for computing quantization weights and local inference; the skill does not include capabilities for network exfiltration or arbitrary command execution based on data content.
  • Sanitization: No specific text sanitization is performed, which is expected for this specific machine learning optimization use case.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 07:53 PM
Security Audit — agent-trust-hub — gptq