Brainstorming Ideas Into Designs
Overview
Help turn ideas into fully formed designs and specs through natural collaborative dialogue, then automatically transition to implementation planning with intelligent subagent orchestration.
Start by understanding the current project context, then ask questions one at a time to refine the idea. Once you understand what you're building, present the design in small sections (200-300 words), checking after each section whether it looks right so far.
Key features:
- •Auto-invokes planning-pipeline after design completion
- •Defines task-to-agent mappings for smart execution
- •Subagent-driven-development uses mappings to dispatch appropriate specialists
The Process
Understanding the idea:
- •Check out the current project state first (files, docs, recent commits)
- •Ask questions one at a time to refine the idea
- •Prefer multiple choice questions when possible, but open-ended is fine too
- •Only one question per message - if a topic needs more exploration, break it into multiple questions
- •Focus on understanding: purpose, constraints, success criteria
Exploring approaches:
- •Propose 2-3 different approaches with trade-offs
- •Present options conversationally with your recommendation and reasoning
- •Lead with your recommended option and explain why
Presenting the design:
- •Once you believe you understand what you're building, present the design
- •Break it into sections of 200-300 words
- •Ask after each section whether it looks right so far
- •Cover: architecture, components, data flow, error handling, testing
- •Be ready to go back and clarify if something doesn't make sense
After the Design (Automatic Pipeline)
Step 1: Documentation with Agent Mapping
Write the validated design to docs/plans/YYYY-MM-DD-<topic>-design.md.
CRITICAL: Include Agent Mapping Section at the end of the design:
markdown
## Agent Mapping for Execution When using subagent-driven-development, dispatch these specialized agents: | Task Pattern | Subagent Type | Model | Notes | |--------------|---------------|-------|-------| | Database/migration tasks | backend-architect | sonnet | Schema, RLS, triggers | | API endpoint tasks | backend-architect | sonnet | Routes, validation, auth | | React component tasks | frontend-developer | sonnet | Components, hooks, state | | TypeScript type tasks | typescript-pro | haiku | Types, interfaces, schemas | | Test writing tasks | general-purpose | sonnet | Unit, integration tests | | Performance tasks | performance-engineer | sonnet | Profiling, optimization | | UI/UX design tasks | ui-ux-designer | sonnet | Design system, accessibility | | Code review tasks | superpowers:code-reviewer | sonnet | Quality gates | ### Task-Specific Agents for This Feature [List specific task → agent mappings based on the design] Example: - Task 1 (Create qc_rules table) → backend-architect - Task 2 (Add QC API endpoints) → backend-architect - Task 3 (Build QCRulesTable component) → frontend-developer - Task 4 (Add Zod schemas) → typescript-pro - Task 5 (Write integration tests) → general-purpose
Use elements-of-style:writing-clearly-and-concisely skill if available. Commit the design document to git.
Step 2: Auto-invoke Planning Pipeline
Once the design document is committed, automatically invoke planning-pipeline:
code
Skill tool -> skill: planning-pipeline
Announce: "Design complete. Now invoking planning-pipeline to create the implementation plan."
The planning-pipeline will:
- •Assess scope (OpenSpec vs Quick-plan path)
- •Create plan artifacts (using writing-plans if Quick-plan path)
- •Preserve agent mappings from design in the plan
- •Get user approval
- •Ingest tasks into Beads with dependencies
- •Offer execution method choice
Do NOT ask "Ready to set up for implementation?" - proceed automatically.
Subagent Orchestration During Execution
When planning-pipeline hands off to subagent-driven-development, the execution follows this enhanced flow:
code
For each task in plan:
1. Read task description
2. Match task to agent type (from agent mapping table)
3. Dispatch appropriate specialist subagent:
Task tool -> subagent_type: [matched-agent-type], model: [sonnet|haiku]
Prompt: "[task description with full context]"
4. Two-stage review:
- Spec compliance review (superpowers:code-reviewer)
- Code quality review (superpowers:code-reviewer)
5. Mark task complete, move to next
Agent Selection Logic
code
Pattern Matching (in priority order): 1. Exact match in task-specific mapping table → Use specified agent 2. Pattern match on task keywords: - "database", "table", "migration", "RLS" → backend-architect - "API", "endpoint", "route", "server action" → backend-architect - "component", "React", "UI", "hook" → frontend-developer - "type", "interface", "schema", "Zod" → typescript-pro - "test", "spec", "e2e" → general-purpose - "performance", "optimize", "cache" → performance-engineer - "design", "UX", "accessibility" → ui-ux-designer 3. Default fallback → general-purpose (sonnet)
Model Selection
code
Task Complexity → Model: - Simple tasks (types, small fixes) → haiku (fast, cheap) - Standard tasks (components, endpoints) → sonnet (quality) - Complex tasks (architecture, security) → sonnet (thoroughness) - Review tasks → sonnet (always)
Parallel Execution
When tasks are independent (no dependency chain):
code
Dispatch multiple subagents in single message:
Task tool (parallel) → [
{ subagent_type: frontend-developer, task: "Build ComponentA" },
{ subagent_type: frontend-developer, task: "Build ComponentB" },
{ subagent_type: typescript-pro, task: "Add types for ComponentA" }
]
Rules for parallelization:
- •No shared file modifications
- •No dependency between tasks
- •Same or compatible domains
- •Max 3 parallel subagents (context management)
Workflow Diagram
code
Complete Flow with Agent Orchestration:
Understand Idea
|
v
Explore Approaches (2-3 options)
|
v
Present Design (sections)
|
v
Write design.md with Agent Mapping Table
|
v
git commit
|
v
[AUTO] Invoke planning-pipeline
|
v
Create plan.md (preserves agent mappings)
|
v
User approves plan
|
v
Beads ingestion (tasks with agent hints)
|
v
User selects execution method
|
v
[If Subagent-Driven]
|
v
For each task:
┌─────────────────────────────────────┐
│ 1. Match task → agent type │
│ 2. Dispatch specialist subagent │
│ 3. Subagent implements + tests │
│ 4. Spec compliance review │
│ 5. Code quality review │
│ 6. Mark complete, next task │
└─────────────────────────────────────┘
|
v
All tasks complete
|
v
Final review + finish branch
Key Principles
- •One question at a time - Don't overwhelm with multiple questions
- •Multiple choice preferred - Easier to answer than open-ended when possible
- •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
- •Seamless handoff - Auto-invoke pipeline without asking
- •Smart agent selection - Match tasks to specialist subagents
- •Parallel when safe - Run independent tasks concurrently
- •Two-stage review - Spec compliance + code quality gates
Skip Planning Option
If user explicitly says they only want the design (not implementation), skip the auto-invoke:
- •"Just the design, no implementation plan"
- •"Design only"
- •"Skip planning"
In these cases, end after committing the design document.
Example: Agent Mapping in Practice
markdown
## Agent Mapping for Execution | Task | Agent Type | Model | Rationale | |------|------------|-------|-----------| | Create qc_rules table | backend-architect | sonnet | DB schema + RLS | | Create qc_results table | backend-architect | sonnet | DB schema + RLS | | Add QC validation RPC | backend-architect | sonnet | Complex SQL logic | | Build QCRulesForm | frontend-developer | sonnet | React + form handling | | Build QCResultsChart | frontend-developer | sonnet | React + Recharts | | Add QC Zod schemas | typescript-pro | haiku | Type definitions | | Write QC unit tests | general-purpose | sonnet | Test coverage | | Write QC e2e tests | general-purpose | sonnet | Integration testing | ### Parallelization Groups - Group 1 (parallel): qc_rules table, qc_results table - Group 2 (sequential, depends on Group 1): QC validation RPC - Group 3 (parallel): QCRulesForm, QCResultsChart, QC Zod schemas - Group 4 (sequential, depends on all): Unit tests, e2e tests
Integration Points
| Skill | Relationship |
|---|---|
| planning-pipeline | Receives design with agent mappings |
| writing-plans | Creates tasks with agent hints |
| subagent-driven-development | Uses mappings to dispatch specialists |
| superpowers:code-reviewer | Two-stage review after each task |
| superpowers:finishing-a-development-branch | Final integration |