lambda-labs-gpu-cloud

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

• Third-party content exposure detected (high risk: 0.90). The skill's required workflows explicitly fetch and execute content from untrusted public sources—e.g., "git clone https://github.com/user/project" in the Training workflows, transformers.from_pretrained('meta-llama/Llama-2-7b-hf') in the fine-tuning workflow, and "aws s3 sync s3://my-bucket/dataset" in Advanced Filesystem Usage—so third-party/user-generated content can be ingested and materially influence subsequent actions.