llama-cpp
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- EXTERNAL_DOWNLOADS (SAFE): The skill references the official GitHub repository for llama.cpp and provides instructions to download model files from HuggingFace. These sources are reputable and the downloads are essential for the skill's stated purpose.
- COMMAND_EXECUTION (SAFE): The skill provides standard build commands (make) and execution commands for the compiled binaries (./llama-cli, ./llama-server). All commands are transparent and match the intended use case of running local LLM inference.
- INDIRECT_PROMPT_INJECTION (SAFE): While the skill involves processing user-provided prompts for inference, it does not possess capabilities that would allow an attacker to perform secondary malicious actions (like file system modification or network exfiltration) through injected prompt content.
Audit Metadata