PACE 最早的可选 stage。想法模糊、写不出验收标准时触发。一次一问 + 推荐答 + 先查库再问人。
Scanned 9/5/2026
Install to Claude Code
npx -y skills add WenJunDuan/Rlues --skill brainstorm --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Brainstorm?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/wenjunduan-brainstorm-16c074e0)More formats (shields.io, HTML) on the badges page.
---
name: brainstorm
description: PACE 最早的可选 stage。想法模糊、写不出验收标准时触发。一次一问 + 推荐答 + 先查库再问人。
---
# /brainstorm — PACE 分诊层 (v9.9.6 · grill-me 化)
## 触发
| 信号 | 来源 | 是否进 brainstorm |
|---|---|---|
| 用户显式 "想法不清楚" / "先 brainstorm" / "帮我想清楚" | 用户 | ✅ |
| 语义模糊: 主 agent 无法从输入直接写出可观测验收标准 (铁律[分诊]) | 自动 | ✅ |
| 用户带方案但说 "听听别的意见" / "pressure-test 一下" | 用户 | ✅ |
| 用户显式 `--skip-brainstorm` / 输入开头 "直接做:" | 用户 | ❌ |
| 显式 bug / 重构描述 | 自动 | ❌ (走 Bugfix/Refactor 路径) |
| ≥ 2 个可独立验收、可独立 ship 的切片且方向已清晰 | 自动 | ❌ (直接进 roadmap) |
| 想法清晰 + 单模块 | 自动 | ❌ (直接 plan) |
## 核心循环 (grill-me)
1. **一次只问一个问题**, 并附上你的**推荐答案** — 用户对草案说 "对/不对/改成X" 远比面对空白提问轻松 (选择题, 不是作文题)
2. **追刚拿到的答案, 再横向换题** — 提前收口都是因为换题太快; 深度来自把一条线问到底
3. **能自己查的不问用户**: 答案在代码/.ai_state/architecture/compound 里 → Grep/Read 自行确认, 连续解决掉的问题只汇报一句
4. **顶住模糊回答**: "以后再说" / "大概X吧" / "都行" = 继续钻的信号, 不是过关信号; 允许指出矛盾与含糊, 礼貌但不接受雾
5. **半答案给 strawman**: 用户说 "不知道, 可能X" → 给一个可反对的完整草案 "这样定, 哪里不对?" — 反对比发明容易
6. **觉得问够了, 再问三个** — "已经够了"的感觉是水面, 不是水底
7. **禁止用总结推进**: "所以你的意思是X/Y/Z" 是收口不是推进; 综合留给最后的 log
## 例外
- 用户明确说 "去查 X 的资料" → 主 agent 可调度 docs_researcher / WebSearch (显式调研)
- 用户说 "看下我们之前怎么决定 Y" → 主 agent 可读 compound/decision-*.md
## 详细 playbook
完整工作流、模板、schema 与联动细节见 `references/playbook.md` —— 按需 Read, 不进热路径。
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!
Ultra-compressed commit message generator. Cuts noise from commit messages while preserving intent and reasoning. Conventional Commits format. Subject ≤50 chars, body only when "why" isn't obvious. Use when user says "write a commit", "commit message", "generate commit", "/commit", or invokes /caveman-commit. Auto-triggers when staging changes.
Ultra-compressed code review comments. Cuts noise from PR feedback while preserving the actionable signal. Each comment is one line: location, problem, fix. Use when user says "review this PR", "code review", "review the diff", "/review", or invokes /caveman-review. Auto-triggers when reviewing pull requests.
一个全面的 Claude Code 会话验证系统。
Verification loop for Django projects: migrations, linting, tests with coverage, security scans, and deployment readiness checks before release or PR.
Verification loop for Spring Boot projects: build, static analysis, tests with coverage, security scans, and diff review before release or PR.