Token导航 LogoToken导航

Agent Skill · 研究检索

think(think 搜索)

think 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

think 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为权限未标注;许可证信息还不明确,第三方扫描有未通过项,按安装量在研究检索类的 73,675 个 Skill 里排第 71,089。

总安装 196周安装 8GitHub Stars 公开资料未说明下载 63核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:think(think 搜索)
来源仓库:https://github.com/fheinonen/think
仓库路径:skills/think
安装命令:
npx skills add https://github.com/fheinonen/think --skill think
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/fheinonen/think --skill think

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

think 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中快速定位候选结果。适用于需要根据关键词或任务场景进行信息定位的场景。通过关键词搜索和来源线索筛选目标信息。

安装命令:npx skills add https://github.com/fheinonen/think --skill think。建议确认权限范围和维护状态后再使用。

来源仓库内路径:skills/think

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/fheinonen/think --skill think;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    权限没有明确标注,第一次使用时只给一个无关紧要的任务,观察它到底读了哪些文件和网络。

常见问题

6 条

think 怎么安装?

在终端执行 npx skills add https://github.com/fheinonen/think --skill think。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

think 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

think 需要什么权限?

权限级别没有明确标注。第一次使用时建议只给无关紧要的任务,观察它实际读写了哪些内容。

think 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

think 用的人多吗?

公开数据是累计安装 196 次,近一周安装 8 次,下载 63 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

think 安全吗?

第三方扫描有未通过项,风险较高。除非你确认过源码,否则不建议直接安装到常用环境。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
think

think 搜索 · 本页

196—权限未标注未明确未通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Think — Thought Organizer

Help the user organize complex thinking by separating capture from structure. The user dumps raw thoughts fast. You organize them into a structured breakdown. Sessions are saved as portable markdown files in ~/.thinking-sessions/.

Modes

Determine the mode from the user's request. If unclear, default to New Session.

New Session

The user wants to think through something new.

Step 1 — Capture. Ask the user:

What are you thinking through? Dump everything in one go. Fragments, sentences, bullet points, whatever comes to mind.

Collect the user's full response. Do not interrupt, ask clarifying questions, or organize anything yet. This is their space.

If the response is empty or very short (under ~20 words), ask again: "I didn't get enough to work with. Try dumping your thoughts again."

Step 2 — Organize. Read all the captured thoughts and produce a Breakdown:

  • Problem — one-sentence framing of what the user is thinking through
  • Threads — numbered list of distinct threads. Each thread gets a bold title and a 1-2 sentence description. Group related thoughts into the same thread.
  • Open Questions — gaps you spotted, things the user hasn't addressed, considerations worth raising. These are things the AI identifies, not just questions the user explicitly asked.
  • Action Items — concrete next steps extracted from the thinking. Format as a checkbox list (- []...).

Present the Breakdown inline.

Step 3 — Generate metadata. From the content, generate:

  • Title: a short descriptive title for the session (used in frontmatter and filename)
  • Slug: 3-5 words, lowercase, hyphenated, derived from the problem statement
  • Tags: 2-5 lowercase hyphenated tags from the domain, client/project context, and topic

Step 4 — Refine. After presenting the Breakdown, ask:

Want to refine? You can: - Add more thoughts - Ask me to dig deeper on a specific thread - Request a different framing - Say "save" to finalize

If the user adds more thoughts: append them to the Raw Thoughts section, re-generate the full Breakdown, and present the updated version.

If the user asks to expand a thread: elaborate on that thread with more detail, implications, and questions. Replace the thread's 1-2 sentence description in the Breakdown with the expanded version (persistent, not ephemeral).

If the user says "save" or "done": proceed to Step 5.

Repeat Step 4 until the user says save or done.

Step 5 — Save. Ensure the directory ~/.thinking-sessions/ exists (create it if not).

Write the session file to ~/.thinking-sessions/YYYY-MM-DD-<slug>.md where YYYY-MM-DD is today's date and <slug> is from Step 3. If a file with that name already exists, append -2 (or -3, etc.) to the slug to avoid collision.

The file format:

---
title: "<title from Step 3>"
created: <current ISO 8601 timestamp>
updated: <current ISO 8601 timestamp>
status: active
tags: [cloud-migration, rollback, infrastructure]
---

## Raw Thoughts

<each thought as a bullet point, preserving the user's original wording>

## Breakdown

### Problem
<problem statement>

### Threads
<numbered thread list>

### Open Questions
<bullet list>

### Action Items
<checkbox list>

Tell the user: "Session saved to ~/.thinking-sessions/<filename>."

List Sessions

The user wants to see their past thinking sessions.

List all .md files in ~/.thinking-sessions/. For each file, read the YAML frontmatter. Only include files that have both title and status fields (skip non-session files).

Sort by filename (newest first). Show max 20 entries.

Output format, one line per session:

YYYY-MM-DD  [status]  title

Example:

2026-04-11  active    Cloud migration rollback strategy
2026-04-10  resolved  Auth system redesign
2026-04-09  active    Client X onboarding flow

If the directory doesn't exist or is empty, say: "No thinking sessions yet. Start one by saying 'I need to think through something.'"

Continue Session

The user wants to continue an existing session.

Find the session file matching the user's description. Search filenames in ~/.thinking-sessions/ for a partial match on the slug.

  • If exactly one file matches: load it.
  • If multiple files match: list them and ask the user to pick.
  • If no files match: say "No session found matching that. Run 'list sessions' to see available sessions."

Once loaded:

  1. Read the file and present the existing Breakdown to the user.
  2. Enter the Capture phase (Step 1 from New Session): ask for more thoughts.
  3. Append new thoughts to the Raw Thoughts section.
  4. Re-generate the full Breakdown incorporating all thoughts (old and new).
  5. Enter the Refine phase (Step 4 from New Session).
  6. On save: overwrite the file with updated content. Update the updated timestamp in frontmatter.

Resolve Session

The user wants to mark a session as done.

Find the session file (same matching logic as Continue Session).

Once found:

  1. If the session is already status: resolved, tell the user it's already resolved and take no action.
  2. Update the frontmatter: set status: resolved and add resolved: <current ISO 8601 timestamp>.
  3. Update the updated timestamp.
  4. Save the file.
  5. Tell the user: "Session '' marked as resolved."

平台分布

Codex

37.5%
按下载量换算24

Claude

32.36%
按下载量换算20

Cursor

18.21%
按下载量换算11

Gemini CLI

8.65%
按下载量换算5

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

未通过

权限和风险

权限未标注

当前来源未能明确判断权限范围,默认进入异常复核队列。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills