lab-review

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly queries and ingests results from public third-party sources—PubMed and medRxiv—via the review-marker script and spawned sub-agents, meaning the agent reads and interprets external article/preprint content (including unreviewed medRxiv abstracts) as part of its workflow, exposing it to untrusted third-party input.