aris-training-check

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

• Third-party content exposure detected (high risk: 0.90). This skill's SKILL.md explicitly reads WandB runs via wandb.Api().run("//<run_id>") and run.history() (and falls back to tailing remote training logs via SSH) and then uses those third-party, user-provided metrics to drive decisions (including killing training), so untrusted external content can directly influence agent actions.