account-scoring

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

• Third-party content exposure detected (medium risk: 0.30). The required runtime workflow (plays/score-accounts.ts -> scoreAccount -> uses.accountScorer(...)) passes operator/account-derived fields (input.name, input.domain) as free text into the LLM agent, and then writes the LLM output back to HubSpot, so an outsider can inject via the account feed/model fields the workflow scores.