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 artifacts guide (references/artifacts.md) explicitly shows adding HTTP/S3 references (e.g., artifact.add_reference('https://example.com/data.zip') and dataset.add_reference('s3://bucket/data')) and then using run.use_artifact(...); artifact.download() to load data and train models, which clearly instructs fetching and ingesting external, potentially untrusted third-party content that can materially change subsequent behavior.