consult

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly invokes external AI providers (gemini, claude, codex, opencode, copilot, kiro) via CLI/ACP and ingests their responses into the workflow (see Provider Configurations, Command Building, ACP Transport, and Output Format), so untrusted third‑party model outputs can be read/interpreted and materially influence subsequent decisions or tool use.