brainstorming（头脑风暴）

Brainstorming Ideas Into Designs

Turn rough ideas into implementation-ready designs through structured collaborative dialogue using Superpower Loop for continuous iteration.

CRITICAL: First Action - Start Superpower Loop NOW

THIS MUST BE YOUR FIRST ACTION. Do NOT explore codebase, do NOT ask questions, do NOT do anything else until you have started the Superpower Loop.

1. Capture $ARGUMENTS as the initial prompt
2. Immediately run:

"${CLAUDE_PLUGIN_ROOT}/scripts/setup-superpower-loop.sh" "Brainstorm: $ARGUMENTS. Continue progressing through the superpowers:brainstorming skill phases: Phase 1 (Discovery) → Phase 2 (Option Analysis) → Phase 3 (Design Creation) → Phase 4 (Design Reflection) → Phase 5 (Git Commit) → Phase 6 (Transition)." --completion-promise "BRAINSTORMING_COMPLETE" --max-iterations 50

1. Only after the loop is running, proceed to explore the codebase and continue with Phase 1

The loop enables self-referential iteration throughout the brainstorming process.

Superpower Loop Integration

This skill uses Superpower Loop to enable self-referential iteration throughout the brainstorming process.

CRITICAL: Throughout the process, you MUST output <promise>BRAINSTORMING_COMPLETE</promise> only when:

• Phase 1-4 (Discovery, Option Analysis, Design Creation, Design Reflection) are all complete
• Design folder created with all required documents
• User approval received in Phase 2
• Git commit completed

Do NOT output the promise until ALL conditions are genuinely TRUE.

ABSOLUTE LAST OUTPUT RULE: The promise tag MUST be the very last text you output. Output any transition messages or instructions to the user BEFORE the promise tag. Nothing may follow <promise>BRAINSTORMING_COMPLETE</promise>.

Initialization

(The Superpower Loop was already started in the critical first action above - do NOT start it again)

1. Context Check: Ensure you have read CLAUDE.md and README.md to understand project constraints.
2. Codebase Index: Verify you have access to the codebase and can run searches.

The loop will continue through all phases until <promise>BRAINSTORMING_COMPLETE</promise> is output.

Core Principles

1. Converge in Order: Clarify → Compare → Choose → Design → Reflect → Commit → Transition
2. Context First: Explore codebase before asking questions
3. Incremental Validation: Validate each phase before proceeding
4. YAGNI Ruthlessly: Only include what's explicitly needed
5. Test-First Mindset: Always include BDD specifications - load superpowers:behavior-driven-development skill

Phase 1: Discovery

Explore codebase first, then ask focused questions to clarify requirements.

Actions:

1. Explore codebase - Use Read/Grep/Glob to find relevant files and patterns
2. Review context - Check docs/, README.md, CLAUDE.md, recent commits
3. Identify gaps - Determine what's unclear from codebase alone
4. Ask questions - Use AskUserQuestion with exactly 1 question per call

- Prefer multiple choice (2-4 options) - Ask one at a time, never bundle - Base on exploration gaps

Open-Ended Problem Context:

If the problem appears open-ended, ambiguous, or requires challenging assumptions:

• Consider applying first-principles thinking to identify the fundamental value proposition
• Question "why" repeatedly to reach core truths
• Be prepared to explicitly load superpowers:build-like-iphone-team skill in Phase 2 for radical innovation approaches

Complexity Assessment (determines sub-agent strategy for Phase 3 & 4):

At the end of discovery, classify the scope:

• Simple (single file/component, clear pattern to follow): Skip sub-agents in Phase 3 & 4 — main agent handles directly
• Medium (cross-module, some architectural decisions): 2 sub-agents in Phase 3 (Architecture + BDD merged, Context), 2 reflection sub-agents in Phase 4
• Complex (new system, large refactor, multiple integration points): Full 3+ sub-agents in Phase 3 and Phase 4

State the assessment explicitly before proceeding.

Output: Clear requirements, constraints, success criteria, relevant patterns, and complexity assessment.

See ./references/discovery.md for detailed patterns and question guidelines. See ./references/exit-criteria.md for Phase 1 validation checklist.

Phase 2: Option Analysis

Research existing patterns, propose viable options, and get user approval.

Actions:

1. Research - Search codebase for similar implementations
2. Identify options - Propose 2-3 grounded in codebase reality, or explain "No Alternatives"
3. Present - Write conversationally, lead with recommended option, explain trade-offs
4. Get approval - Use AskUserQuestion, ask one question at a time until clear

Radical Innovation Context:

If the problem involves:

• Challenging industry conventions or "how things are usually done"
• Creating a new product category rather than improving existing
• Questioning fundamental assumptions
• Open-ended or ambiguous requirements that need disruptive thinking

Then explicitly load superpowers:build-like-iphone-team skill using the Skill tool to apply iPhone design philosophy (first-principles thinking, breakthrough technology, experience-driven specs, internal competition, Purple Dorm isolation).

Output: User-approved approach with rationale and trade-offs understood.

See ./references/options.md for comparison and presentation patterns. See ./references/exit-criteria.md for Phase 2 validation checklist.

Phase 3: Design Creation

Create design documents. Scale sub-agent usage based on complexity assessment from Phase 1.

Simple: Main agent handles all research and document creation directly. No sub-agents needed — explore codebase, search for best practices, write BDD scenarios, and create design documents in a single pass.

Medium (2 sub-agents):

Sub-agent 1: Architecture & Best Practices Research

• Focus: Existing patterns, libraries, WebSearch for best practices, security, performance
• Load superpowers:behavior-driven-development skill
• Output: Architecture recommendations, BDD scenarios, best practices summary

Sub-agent 2: Context & Requirements Synthesis

• Focus: Synthesize Phase 1 and Phase 2 results
• Output: Context summary, requirements list, success criteria

Complex (3+ sub-agents):

Sub-agent 1: Architecture Research

• Focus: Existing patterns, architecture, libraries in codebase
• Use WebSearch for latest best practices
• Output: Architecture recommendations with codebase references

Sub-agent 2: Best Practices Research

• Focus: Web search for best practices, security, performance patterns
• Load superpowers:behavior-driven-development skill
• Output: BDD scenarios, testing strategy, best practices summary

Sub-agent 3: Context & Requirements Synthesis

• Focus: Synthesize Phase 1 and Phase 2 results
• Output: Context summary, requirements list, success criteria

Additional sub-agents: Launch for distinct, research-intensive aspects as needed.

Integrate results: Merge all findings, resolve conflicts, create unified design.

Design document structure:

docs/plans/YYYY-MM-DD-<topic>-design/
├── _index.md              # Context, Requirements, Rationale, Detailed Design, Design Documents section (MANDATORY)
├── bdd-specs.md           # BDD specifications (MANDATORY)
├── architecture.md        # Architecture details (MANDATORY)
├── best-practices.md      # Best practices and considerations (MANDATORY)
├── decisions/             # ADRs (optional)
└── diagrams/              # Visual artifacts (optional)

CRITICAL: _index.md MUST include Design Documents section with references:

## Design Documents

- [BDD Specifications](./bdd-specs.md) - Behavior scenarios and testing strategy
- [Architecture](./architecture.md) - System architecture and component details
- [Best Practices](./best-practices.md) - Security, performance, and code quality guidelines

Output: Design folder created with all files saved.

See ./references/design-creation.md for sub-agent patterns and integration workflow. See ./references/exit-criteria.md for Phase 3 validation checklist.

Phase 4: Design Reflection

Before committing, verify design quality. Scale reflection based on complexity assessment from Phase 1.

Simple: Main agent performs a single review pass — check requirements coverage, BDD completeness, and document consistency sequentially. No sub-agents needed.

Medium (2 reflection sub-agents):

Sub-agent 1: Requirements & BDD Review

• Focus: Verify requirements traceability AND BDD scenario completeness (happy path, edge cases, errors)
• Output: Traceability matrix, missing scenarios, coverage gaps

Sub-agent 2: Consistency & Risk Review

• Focus: Cross-document terminology, references, component names, and key risks
• Output: Inconsistencies, broken references, unaddressed risks

Complex (3+ reflection sub-agents):

Sub-agent 1: Requirements Traceability Review

• Focus: Verify every Phase 1 requirement is addressed in design
• Output: Traceability matrix, orphaned requirements list

Sub-agent 2: BDD Completeness Review

• Focus: Check BDD scenarios cover happy path, edge cases, and error conditions
• Output: Missing scenarios list, coverage gaps

Sub-agent 3: Cross-Document Consistency Review

• Focus: Verify terminology, references, and component names are consistent
• Output: Inconsistencies list, terminology conflicts

Additional sub-agents (launch as needed): Security Review, Risk Assessment.

Integrate and Update:

1. Collect all sub-agent findings
2. Prioritize issues by impact
3. Update design documents to fix issues
4. Re-verify updated sections
5. Confirm with user: Use AskUserQuestion to present the reflection summary and get approval before committing

Output: Updated design documents with issues resolved and user approval received.

See ./references/reflection.md for sub-agent prompts and integration workflow.

Phase 5: Git Commit

Commit the design folder using git-agent (with git fallback).

Actions:

1. Stage the entire folder: git add docs/plans/YYYY-MM-DD-<topic>-design/
2. Run: git-agent commit --no-stage --intent "add design for <topic>" --co-author "Claude <Model> <Version> <noreply@anthropic.com>"
3. On auth error, retry with --free flag
4. Fallback: If git-agent is unavailable or fails, use git commit -m "docs: add design for <topic>..." with conventional format

See ../../skills/references/git-commit.md for detailed patterns.

Phase 6: Transition to Implementation

Prompt the user to use superpowers:writing-plans, then output the promise as the absolute last line.

Output in this exact order:

1. Transition message: "Design complete. To create a detailed implementation plan, use /superpowers:writing-plans."
2. <promise>BRAINSTORMING_COMPLETE</promise> — nothing after this

PROHIBITED: Do NOT offer to start implementation directly. Do NOT output any text after the promise tag.

References

• ./references/core-principles.md - Core principles guiding the workflow
• ./references/discovery.md - Exploration patterns and question guidelines
• ./references/options.md - Option comparison and presentation patterns
• ./references/design-creation.md - Sub-agent patterns, integration workflow, design structure
• ./references/reflection.md - Design reflection patterns and gap identification strategies
• ./references/exit-criteria.md - Validation checklists, success indicators, common pitfalls
• ../../skills/references/git-commit.md - Git commit patterns and requirements (shared cross-skill resource)
• ../../skills/references/loop-patterns.md - Completion promise design, prompt patterns, and safety nets