The Agent Skills Directory

[SAFE]: The skill provides instructional content and code snippets for LLM fine-tuning using established industry-standard libraries such as transformers, peft, and vllm. All suggested practices align with secure machine learning development workflows.
[DATA_EXFILTRATION]: No hardcoded secrets or sensitive file path access were detected. The examples correctly demonstrate using environment variables or local server configurations for API clients.
[PROMPT_INJECTION]: While the skill is designed to process external training data, it includes robust mechanisms for data validation and quality filtering to prevent the ingestion of malicious or low-quality instructions.
[COMMAND_EXECUTION]: Shell command usage is confined to legitimate model conversion and deployment tasks, such as quantizing models with llama.cpp tools or starting vLLM inference servers.

fine-tuning-expert