AI Agent Hub
Back to skills
Kimi Web Search icon

Kimi Web Search

Knowledge Management Updated 2026.08.30

Paste the following prompt into your AI chat to install this skill:

Please follow https://skillhub.cn/install/skillhub.md to install @redfox-data/kimi-websearch-redfox.

About this skill

Problem

When an agent needs up-to-date web information, direct model calls can be unstable, make waiting semantics unclear, and leave citations scattered. Kimi web search separates the task into one submission and continuous polling, so the caller consumes structured JSON instead of handling endpoint plumbing.

How it works

The skill wraps the async Kimi WebSearch API and usually runs in three stages:
- Extract keywords: derive inquiry_text from natural language while preserving the user's actual search intent.
- Submit and poll: call POST /story/api/kimi/submit, then query POST /story/api/kimi/result every 5 seconds for up to 5 minutes; progress goes to stderr, final JSON goes to stdout.
- Present structured output: after status becomes completed, read Kimi's answer from data.content, cited sources from data.result, and format them for the user.

Boundaries

It fits real-time web retrieval where cited sources matter and async waiting is acceptable, not offline knowledge lookup or low-latency requests that must return immediately. It requires REDFOX_API_KEY, and callers should handle submission errors, polling timeouts, and task failed states.

Use Cases

  • Compile competitor release notes and user complaints into a cited summary during product research.
  • During support review, search current error reports and known issues for a feature, then generate troubleshooting points.
  • Before a weekly market report, retrieve industry news, policy updates, and competitor moves from the last 48 hours.
  • During tech selection, search recent security notices, breaking changes, and maintenance status of frameworks, then summarize the comparison.

Best For

  • Product managers tracking competitors: need weekly release updates, announcements, and user feedback with cited sources.
  • Support engineers reviewing incidents: need quick retrieval of error symptoms, known issues, and workarounds.
  • Analysts writing industry briefs: need time-bounded collection of news, policies, and competitor moves with sources.
  • Architects choosing technology: need framework security advisories, breaking changes, and maintenance signals for comparison.