kimi-websearch

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

• Third-party content exposure detected (high risk: 0.75). Required workflow calls scripts/kimi_search.py, which fetches runtime search results from https://redfox.hk/story/api/kimi/result and then prints result_data (including Kimi answer text and citations) as JSON to stdout; that returned free text is then consumed by the agent, creating an indirect prompt-injection path from outsider-provided web search content.