AgentSkillsCN

brainstorm

为长时间的头脑风暴提供深度支持。在开展任何创意工作之前使用:无论是功能设计、组件开发、架构规划,还是战略制定。融合协作对话(每次只聚焦一个问题)、系统思考(参与者/激励机制/二阶效应),以及25种跨领域心智模型。最终产出经过验证的设计方案,可用于文档撰写或方案规划。

SKILL.md
--- frontmatter
name: brainstorm
description: >
  Deep brainstorming for long sessions. Use before any creative work: features, components, architecture, strategy.
  Combines: collaborative dialogue (one question at a time), systems thinking (players/incentives/2nd-order effects),
  and 25 cross-domain mental models. Outputs validated designs to docs/plans/.

Deep Brainstorming

Turn ideas into fully-formed designs through collaborative dialogue, systems analysis, and cross-domain thinking.

Announce at start: "Starting brainstorm session. I'll ask one question at a time."

The Process

Phase 1: Understand the Idea

  1. Check current project context (files, docs, recent commits)
  2. Ask questions one at a time to refine the idea
  3. Prefer multiple choice when possible
  4. Focus on: purpose, constraints, success criteria, stakeholders

Phase 2: Analyze the System

Before proposing solutions, map the system:

  1. Map players & incentives — Who's involved? What does each want?
  2. Identify stocks & flows — What accumulates? What moves between states?
  3. Trace 2nd-order effects — If we do X, what happens next? Then what?
  4. Find leverage points — Where can small changes create large impact?
  5. Spot feedback loops — What reinforces or balances itself?

Phase 3: Apply Mental Models

Pick 2-3 relevant lenses from the toolbox below to analyze from different angles.

Phase 4: Explore Approaches

  • Propose 2-3 different approaches with trade-offs
  • Lead with your recommendation and explain why
  • Apply YAGNI ruthlessly — remove unnecessary features

Phase 5: Present the Design

  • Break into sections of 200-300 words
  • Ask after each section: "Does this look right so far?"
  • Cover: architecture, components, data flow, error handling, testing
  • Go back and clarify when something doesn't fit

Phase 6: Document

Write validated design to docs/plans/YYYY-MM-DD-<topic>-design.md


Systems Thinking Framework

Core Questions

QuestionPurpose
Who are all the players, and what does each want?Map incentives
If we do X, what happens next? Then what?Trace cascading effects
What accumulates over time? What flows?Model dynamics
Where are the feedback loops?Find reinforcing/balancing forces
What constraint, if removed, unlocks the most value?Find leverage
What recurring pain could be systematized?Identify automation opportunities

Common Mistakes to Flag

  • Only seeing first-order effects — Changes ripple in non-obvious ways
  • Ignoring incentives — Players respond to their incentives, not yours
  • Optimizing locally — Improving one part can worsen the whole
  • Missing feedback loops — Systems amplify or dampen changes
  • Treating symptoms — Address root causes, not visible symptoms

25 Mental Models Toolbox

#LensCore QuestionBest For
1ArtistWhat makes this unique?Creative blocks, innovation
2EconomistHow do people respond to incentives?Behavior prediction
3EngineerCan I model and calculate this?Data-driven decisions
4EntrepreneurWhat works if I try many things?Uncertainty, experimentation
5DoctorWhat's the diagnosis from symptoms?Root cause analysis
6JournalistHave I verified from independent sources?Research, validation
7ScientistDoes this withstand controlled testing?Hypothesis testing
8MathematicianCan I prove this rigorously?Logic, error detection
9ProgrammerWhat patterns can I automate?Process optimization
10ArchitectWhat will this look like at full scale?Future visualization
11SalespersonWhat do people really want beneath stated needs?Understanding motivations
12SoldierWhat procedure must I follow exactly?Risk prevention
13Chess MasterWhat happens next if I simulate this?Strategic foresight
14DesignerDoes this intuitively suggest how to use it?UX, communication
15TeacherHow do I build knowledge in a learner's mind?Explanation, transfer
16AnthropologistCan I understand this group from inside?Culture analysis
17PsychologistDoes my model predict actual behavior?Human behavior
18CriticHow can I build on others' work?Analysis, synthesis
19PhilosopherWhat happens when I push this to extremes?Finding flaws
20AccountantWhat ratios reveal hidden truths?Metrics analysis
21PoliticianWhat will people believe about this?Perception strategy
22NovelistDoes my story make coherent sense?Narrative structure
23ActorCan I actually feel the state I need?Emotional management
24PlumberWhat would I find by examining directly?Hands-on debugging
25HackerWhat's really happening underneath?Deep system understanding

Quick Model Selection

Problem TypeRecommended Lenses
Need creativityArtist, Entrepreneur, Designer
Understanding behaviorEconomist, Psychologist, Salesperson
Making predictionsEngineer, Chess Master, Scientist
Debugging issuesDoctor, Plumber, Hacker
Improving processesProgrammer, Accountant, Architect
CommunicationNovelist, Teacher, Designer
Decision under uncertaintyEntrepreneur, Scientist, Politician
Understanding peopleAnthropologist, Psychologist, Actor
Finding hidden assumptionsPhilosopher, Mathematician, Critic
Risk managementSoldier, Accountant, Engineer

Long Session Discipline

Pacing

  • One question per message — Don't overwhelm
  • Validate incrementally — Check understanding before building on it
  • Take breaks — After major design sections, pause for confirmation
  • Document as you go — Don't rely on memory across long sessions

Staying Grounded

  • Evidence before claims — Don't assume; verify
  • Name your lens — "Looking at this as an Economist..." makes reasoning visible
  • Revisit assumptions — Early assumptions may become invalid as understanding deepens
  • Track open questions — Maintain a list of unresolved items

Session Markers

Use these to structure long sessions:

code
🎯 GOAL: [What we're trying to figure out]
📍 CHECKPOINT: [Summary of where we are]
❓ OPEN: [Unresolved questions]
✅ DECIDED: [Confirmed decisions]
🔄 REVISIT: [Things to reconsider later]

Key Principles

  • One question at a time — Don't overwhelm with multiple questions
  • Multiple choice preferred — Easier to answer than open-ended
  • YAGNI ruthlessly — Remove unnecessary features from all designs
  • Explore alternatives — Always propose 2-3 approaches before settling
  • Incremental validation — Present design in sections, validate each
  • Be flexible — Go back and clarify when something doesn't make sense
  • Guide, don't answer — Use questions to help thinking, not give solutions
  • Combine lenses — Powerful insights come from mixing perspectives

Output Template

markdown
# [Topic] Design

**Date:** YYYY-MM-DD
**Goal:** [One sentence]
**Status:** Draft | Validated

## Context
[Current state, constraints, stakeholders]

## System Map
- **Players:** [Who's involved and their incentives]
- **Stocks:** [What accumulates]
- **Flows:** [What moves between states]
- **Leverage points:** [Where small changes have big impact]

## Approaches Considered
1. [Approach A] — [Trade-offs]
2. [Approach B] — [Trade-offs]
3. [Approach C] — [Trade-offs]

**Chosen:** [Which and why]

## Design
[Architecture, components, data flow]

## Open Questions
- [ ] [Unresolved item 1]
- [ ] [Unresolved item 2]

## Next Steps
1. [Action 1]
2. [Action 2]

Sources

Fused from: