weights-and-biases

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

• Third-party content exposure detected (high risk: 0.80). The skill's workflow explicitly downloads and uses external artifacts (see references/artifacts.md: artifact.add_reference('https://example.com/data.zip') and run.use_artifact('training-data:latest') followed by artifact.download() and load_data(...)), which causes the agent to fetch and process untrusted user-provided third‑party content (S3/GCS/HTTP/artifacts) that can materially influence training/configuration and downstream actions.