review-salience-xlsx

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

• Third-party content exposure detected (high risk: 0.75). The skill explicitly supports swapping the built-in scorer for external webhooks or arbitrary REST APIs (see references/external-scorer.md and Step 3 "Scorer contract"), sending reviews to third‑party endpoints and ingesting their responses which directly determine salience scores and downstream PCA/K-means decisions, so untrusted third‑party content can materially influence the agent's behavior.