gguf-quantization
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches source code from the official llama.cpp GitHub repository and model weights from Hugging Face (meta-llama).
- [COMMAND_EXECUTION]: Provides instructions for building native binaries, managing system resources (ulimit), and executing quantization tools.
- [PRIVILEGE_ESCALATION]: Recommends system-level package installation (sudo apt install) for required GPU drivers and toolkits within the troubleshooting documentation.
- [INDIRECT_PROMPT_INJECTION]: The skill establishes an environment for processing external data. Ingestion points: user prompts and calibration text (SKILL.md, advanced-usage.md). Boundary markers: absent. Capability inventory: subprocess calls (nvidia-smi), file-write (Modelfile, calibration.txt), and network operations (server mode). Sanitization: present (demonstrates LlamaGrammar for constrained JSON output).
Audit Metadata