Token导航 LogoToken导航

Agent Skill · 研究检索

cw-brainstorming(连续波头脑风暴)

cw-brainstorming 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。

cw-brainstorming 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,权限为只读;许可证信息还不明确,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 25,433。

总安装 3,770周安装 154GitHub Stars 147下载 1,220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:cw-brainstorming(连续波头脑风暴)
来源仓库:https://github.com/haowjy/creative-writing-skills
仓库路径:skills/cw-brainstorming
安装命令:
npx skills add https://github.com/haowjy/creative-writing-skills --skill cw-brainstorming
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/haowjy/creative-writing-skills --skill cw-brainstorming

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。

cw-brainstorming 捕获故事构思与创作探索,以工作笔记形式保留模糊选项与灵感标记。适用于 Codex、Claude、Cursor、Gemini CLI 的创意开发阶段,支持世界构建与角色发展。记录时不添加未声明细节,保持未来创作灵活性,所有建议均标明来源。

输出为开放式骨架,需用户后续决策转化,避免过度承诺具体情节走向。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

来源仓库内路径:skills/cw-brainstorming

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/haowjy/creative-writing-skills --skill cw-brai…;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

    它标注为只读,先用只读任务试一次,确认行为符合预期再扩大使用范围。

常见问题

6 条

cw-brainstorming 怎么安装?

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

cw-brainstorming 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。

cw-brainstorming 需要什么权限?

权限级别是只读。该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

cw-brainstorming 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。

cw-brainstorming 用的人多吗?

公开数据是累计安装 3,770 次,近一周安装 154 次,下载 1,220 次,GitHub 147 星。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。

cw-brainstorming 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

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。
  • 标注为只读权限,试用阶段不容易误改本地文件。

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。
  • 只想先试试效果、不希望它改动本地文件的人。

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
cw-brainstorming

连续波头脑风暴 · 本页

3,770147只读未明确有提醒

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

Brainstorming Capture

Capture story brainstorming in working note format that preserves creative freedom.

Core Principle

Record brainstorming WITHOUT:

  • Over-elaborating on what was stated
  • Mixing user statements with AI suggestions unmarked
  • Inventing excessive details
  • Constraining future creativity

AI suggestions are valuable but must be clearly marked and kept minimal.

Types of Brainstorming

This skill handles all brainstorming types:

  • Story/plot directions (general narrative exploration)
  • Chapter structure and beats (planning individual chapters)
  • Worldbuilding and lore (magic systems, cultures, history, geography)
  • Character development (motivations, arcs, relationships)
  • Timeline and continuity (chronology, contradictions)

All share core principles (minimal capture, source tagging, preserve vagueness). See references/ for specialized guidance:

  • chapter-planning.md - Capturing beat and scene exploration
  • worldbuilding.md - Exploring fictional world elements (use web search for research)
  • character-development.md - Exploring motivations, arcs, relationships
  • continuity-timeline.md - Timeline tracking and contradiction handling

Critical Rules

1. Minimal Capture Only

Record ONLY what the user explicitly states. Do NOT add elaborations, examples they didn't give, or details to fill gaps.

The problem is mixing, not suggesting:

❌ User: "Character A competes with B" → Capture: "A and B compete for leadership through a tournament with three rounds..." ✅ User: "Character A competes with B" → Capture: "A and B compete" + optional: "Tournament? Political? Trial?"

2. Source Tagging (Simple 3-Tag System)

Default: Untagged = user said it. Most ideas come from the user, so treat them as the default.

ONLY use tags for special context:

  1. <AI>...</AI> - AI suggestions/possibilities (MUST be clearly wrapped)

- Use when offering ideas user didn't state - Keep to 2-3 brief options - Example: <AI>Competition could be: tournament-style, political maneuvering, or trial-based</AI>

  1. <hidden>...</hidden> - Author-only information meant to be revealed later

- Secret character motivations - Planned twists/revelations - Behind-the-scenes reasons unknown to characters/readers yet - Example: <hidden>Z secretly wants them both to fail so he can reclaim leadership</hidden>

When to offer AI suggestions:

  • User asks for ideas
  • User seems stuck
  • Offering brief possibilities to spark creativity

When to stay minimal:

  • User is actively exploring their own ideas
  • Just capturing an ongoing discussion
  • User didn't ask for suggestions

3. Preserve Vagueness

Keep it vague if user leaves it vague:

  • "might create tension" → Record as uncertain
  • "thinking about" → Record as consideration
  • "maybe" → Record as possibility

4. Multiple Options Coexist

Working notes can contain contradictions and multiple possibilities. Don't resolve them - just list the options being considered.

Output Approach

Use whatever structure fits the discussion. Could be:

  • Bullet lists
  • Sections organized by topic
  • Timeline format
  • Character-focused groupings
  • Whatever captures the brainstorm clearly

Essential elements:

  • Minimal capture (user's words, not elaborations)
  • Vagueness preserved
  • AI suggestions wrapped in <AI> tags
  • Author-only info wrapped in <hidden> tags when relevant

Optional sections based on discussion:

  • Open questions to explore
  • Multiple options being considered
  • AI suggestions (if offered)
  • Contradictions to resolve later

Teaching Example: The Distinction

User Says:

"I'm thinking character X and character Y compete for leadership. Maybe this creates tension with character Z who was the previous leader."

✅ Good Capture:

# Leadership Competition Notes

- X and Y compete for leadership
- Z was previous leader
- May create tension with Z (uncertain)

Open questions:
- Form of competition?
- How does Z respond?
- Outcome?

❌ Bad Capture:

# Leadership Competition Arc

X and Y compete for leadership after Z steps down. Z feels threatened by the challenge to his authority.

The competition unfolds in three stages:
1. Announcement and initial positioning
2. First challenge where X demonstrates strength
3. Second challenge where Y shows wisdom
...
[20 more invented beats]

Why bad? Added massive elaboration the user never stated.

✅ Good with AI Suggestions:

# Leadership Competition Notes

- X and Y compete for leadership
- Z was previous leader
- May create tension with Z (uncertain)

Open questions:
- Competition format: <AI>tournament-style? political maneuvering? trial-based?</AI>
- Z's response: <AI>oppose both? support one? stay neutral?</AI>
- Resolution?

✅ Good with Hidden Author Notes:

# Leadership Competition Notes

- X and Y compete for leadership
- Z was previous leader
- May create tension with Z (uncertain)
- <hidden>Z is secretly manipulating both X and Y to destroy each other, planning to reclaim power after they're both discredited</hidden>

Open questions:
- Competition format?
- Outcome?

Why use <hidden>? The manipulation twist is planned for later reveal. Readers/characters don't know yet, but the author needs to track it while brainstorming.

If You're Over-Elaborating

Stop if you're writing:

  • Numbered scene lists
  • Detailed backstories
  • Specific dialogue
  • Precise timelines
  • Multiple paragraphs per point
  • Examples user didn't mention

Wrap AI suggestions in <AI> tags, keep minimal (2-3 options).

Success Check

Good: User says "Yes, that's what I said" Bad: User says "I never said all that"

Notes should feel skeletal and incomplete. That's the point - preserves creative freedom.

After Capturing: Discuss and Explore

DON'T just write notes and stop. After capturing, engage with the user to help develop ideas:

Useful follow-ups:

  • Clarifying questions: "You mentioned tension with Z - are you thinking internal conflict or external confrontation?"
  • Potential directions: "This setup could go a few ways: political intrigue, personal drama, or action-focused. What feels right?"
  • Exploring implications: "If Z opposes them both, how does that change the power dynamics?"
  • Connecting threads: "This competition ties into the earlier succession crisis you mentioned - want to explore that link?"

Keep it conversational:

  • Offer 2-3 possibilities, not exhaustive lists
  • Ask about what excites the user
  • Help clarify vague ideas without over-defining them
  • Point out interesting implications or contradictions

The goal: Help the user think through their ideas, not take over the creative process.

Skills are Composable

Feel free to combine with other skills when helpful (e.g., using cw-official-docs to document finalized worldbuilding, or cw-story-critique to analyze what you're brainstorming).

File Placement (Claude Code)

  1. Check project docs for conventions
  2. Look at where similar content lives
  3. Place near related content
  4. Name: brainstorm-[topic].md
  5. Ask if unclear

平台分布

Codex

36.42%
按下载量换算444

Claude

32.73%
按下载量换算399

Cursor

19.27%
按下载量换算235

Gemini CLI

9.47%
按下载量换算116

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

可疑

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

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