unsloth-training

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly loads public third-party data (e.g., reference/grpo/basic_grpo.py calls load_dataset("openai/gsm8k") and SKILL.md/vision-training.md show loading HuggingFace datasets and downloading models), so it ingests untrusted public dataset content that the agent is expected to read and use to drive training decisions and actions—creating a clear avenue for indirect prompt injection.