The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The convert_lora_to_gguf.py script clones the official llama.cpp source code from its public GitHub repository (https://github.com/ggerganov/llama.cpp.git) to utilize the required conversion scripts and the gguf-py package.
[COMMAND_EXECUTION]: The skill relies on local execution of the llama.cpp toolchain, including llama-cli, llama-server, and finetune. It uses subprocess.run to orchestrate model conversion, quantization, and server management.
[DATA_EXPOSURE]: The ollama_model_path.sh script accesses the user's .ollama/models directory to resolve model hashes into local file paths. This is a functional requirement to enable direct inference on models already managed by Ollama without data duplication.

llama-cpp