kimi-result-handling

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The prompt requires quoting Kimi verbatim (including commands, code, or any tokens it may emit) and explicitly forbids altering or auto-executing them, which forces the model to reproduce any secrets or API keys Kimi outputs without sanitization — a clear exfiltration risk.

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

• Third-party content exposure detected (high risk: 0.85). The skill is about rendering and quoting the kimi output string (response) that the companion already aggregated from the polycli-companion.bundle.mjs call; that response is outsider-authored model output (from the external Kimi provider) and is inserted verbatim into the agent’s LLM context for presentation.