fine-tuning-expert

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (low risk: 0.10). The skill’s runtime path only describes using user-provided dataset fields (e.g., instruction, input, output / conversations[*].value) in training/evaluation code, not any outsider-posted free-text feed or queue that the workflow automatically reads without selecting a specific dataset/file.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill calls AutoModelForCausalLM.from_pretrained with trust_remote_code=True which causes the model repo code to be fetched and executed at runtime (e.g., meta-llama/Llama-3.1-8B).