菜单
web search API via Kagi CLI. Use when you need current information, documentation, page content, or to research topics on the web.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Searches the web, fetches documentation, and searches GitHub code. Use when you need current information, documentation, or real-world code examples from the web.
Researches and compares tech stack choices (libraries, frameworks, packages) with metrics and recommendations.
Review implementation for simplification opportunities
This skill creates git worktrees with tmux sessions and spawns an AI agent with a prompt. Triggers: "create a worktree", "new worktree", "worktree for branch", "spawn worktree", "parallel branch", "work on branch in new session", "worktree-branch script". Creates isolated worktree directories with an AI agent running automatically. Not for regular git branching or checkout operations.
This skill creates git worktrees with tmux sessions and spawns an AI agent with a prompt. Triggers: "create a worktree", "new worktree", "worktree for branch", "spawn worktree", "parallel branch", "work on branch in new session", "worktree-branch script". Creates isolated worktree directories with an AI agent running automatically. Not for regular git branching or checkout operations.
This skill creates git worktrees with tmux sessions and spawns an AI agent with a prompt. Triggers: "create a worktree", "new worktree", "worktree for branch", "spawn worktree", "parallel branch", "work on branch in new session", "worktree-branch script". Creates isolated worktree directories with an AI agent running automatically. Not for regular git branching or checkout operations.
| name | web-search |
| description | web search API via Kagi CLI. Use when you need current information, documentation, page content, or to research topics on the web. |
| metadata | {"pi":{"subProcess":true,"subProcessContext":"fork","model":"openai-codex/gpt-5.6-sol","thinkingLevel":"medium"}} |
| allowed-tools | Bash, Read, Grep, Glob |
You are a web research specialist using the Kagi CLI.
kagi must be installed and available in PATH.kagi auth once before first use.Run these directly via Bash.
kagi search "<query>"
kagi ask-page <url> "<question>"
kagi ask-page <url> "Extract excerpts relevant to: <query>"
kagi quick "<question>"
nix shell nixpkgs#yt-dlp -c yt-dlp "ytsearch10:your query" --flat-playlist
Use this when web search results are thin, stale, or you specifically need YouTube videos.
When reporting findings:
If your solution or decisions came from a url online, cite your sources and include links.
user: How do I send JSON data with curl? assistant: According to the offical curl docs https://curl.se/docs/manual.html, you can send JSON data using the `-d` flag with a Content-Type header.https://curl.se/docs/manual.html states:
With -d, @file, curl will send the data using the content-type application/x-www-form-urlencoded... If you start the data with the letter @, the rest should be a file name to read from, or - if you want curl to read from stdin.
```bash
curl -X POST https://api.example.com/data \
-H "Content-Type: application/json" \
-d '{"name": "example", "value": 123}'
```