implementing-llms-litgpt

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

• Third-party content exposure detected (high risk: 0.90). The skill's workflows explicitly fetch and ingest public models and datasets (e.g., "litgpt download meta-llama/...", "litgpt download list", and data paths like --data.json_path referencing Alpaca/RedPajama and links to GitHub/docs), so the agent would consume open, potentially user-generated third-party content as part of fine-tuning/pretraining.