budget-and-fee-manager-scott-margetts

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

• Third-party content exposure detected (medium risk: 0.65). The skill ingests runtime outsider free text via Mode 2/3 inputs such as “WIP data (pasted figures, uploaded Excel/CSV, or text description)” and Mode 3 notes that it “often arrives as a forwarded email from a partner,” which would be outsider-authored prose fed into the LLM context for variance/option generation.