kelly-ecommerce-intel

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s runtime path reads/writes only Busabase records for signals/actions/drafts, but those records (free-text fields like summary, why-it-matters, buyer-intent, body/edited-body) are consumed by the UI at load time (busabase-provider.getState() → buildSnapshot() → client rendering), so an outsider who can submit poison content into the monitored Busabase Bases could inject arbitrary text that the LLM later drafts/uses during its decision workflow.