model-pruning

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

• Third-party content exposure detected (high risk: 0.70). The skill explicitly ingests public, user-generated calibration corpora (e.g., load_dataset("allenai/c4") in references/wanda.md and load_dataset("wikitext", ...) in production_pruning_pipeline) and runs the model on that open web text as part of pruning, so it consumes untrusted third-party content that could carry adversarial instructions.