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 includes automated remote job code and workflows that fetch and execute user-provided content from public sources—e.g., git clone/git pull of arbitrary GitHub repos, transformers.from_pretrained(...) (Hugging Face models), aws s3 sync / rclone, and the run_remote_job/monitor_job code that executes commands and tails logs on instances—so the agent can ingest and read untrusted, user-generated third‑party content at runtime.