carbon-accounting

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

• Third-party content exposure detected (high risk: 1.00). The SKILL.md explicitly requires querying and validating external/public data sources—e.g., "Identify external API integrations (utility APIs, EF databases, IoT feeds)" in Phase 1, "Verify emission factor accuracy against EPA AP-42, IPCC, or regional databases" in Scope 1/2, and "Validate retirement status verification against registry APIs (Verra, Gold Standard, ACR, CAR)" in Phase 5—which means the agent will fetch and interpret untrusted third-party web/API content that can materially alter decisions.