Ideation Expertise

Purpose

Ideation is the creative, divergent-thinking phase that happens before Discovery. It captures messy brainstorming — problems, ideas, user needs, business context — and produces a structured document that Discovery can consume.

Key principle: Ideation is tool-agnostic. This skill works in AgentFlow, but the same approach works in Claude web, ChatGPT, workshops, or whiteboards. The output format is what matters, not the tool.

Portable Skill

This skill is designed to be portable. The content below can be copied into any AI tool's system prompt or project instructions. Stakeholders can brainstorm in their preferred environment and produce a document that plugs directly into AgentFlow's Discovery phase.

Ideation Guide

What Ideation IS

• Exploring problems, opportunities, and ideas
• Capturing domain knowledge from stakeholders
• Describing user workflows, screens, and interactions
• Articulating business context and strategic alignment
• Identifying constraints, risks, and open questions

What Ideation is NOT

• Making technical decisions (architecture, data models, APIs)
• Selecting technology stacks or frameworks
• Designing database schemas or system integrations
• Committing to implementation approaches

The Precision Boundary

Ideation can be highly precise about the problem space and user experience. It must stay silent on technical implementation.

If the conversation drifts toward implementation: Redirect with: "That's a great question for Discovery, where we'll have full codebase context. For now, let's capture what the user needs to accomplish."

Guiding Questions

Use these themes to structure brainstorming. Not all questions apply to every project — use what's relevant.

Theme 1: Problem Space

• What problem are we solving? Who has this problem?
• How is this problem being solved today? What's broken about that?
• What happens if we don't solve this? What's the cost of inaction?
• How urgent is this? Is there a deadline or trigger event?

Theme 2: Opportunity Space

• What's the opportunity if we solve this well?
• Who benefits most? (Users, business, operations, partners)
• Are there adjacent opportunities we should be aware of?
• What does success look like in 6 months? 2 years?

Theme 3: User Space

• Who are the users? What are their roles and skill levels?
• Walk me through their current workflow step by step
• Where do they get frustrated? Where do they lose time?
• What would their ideal experience look like?
• Are there different user types with different needs?

Theme 4: Business Context

• Why is the business investing in this now?
• How does this align with company strategy or vision?
• What are the business constraints (budget, timeline, resources)?
• Are there compliance, regulatory, or legal requirements?
• What business metrics should this move?

Theme 5: Ideas and Features

• What are the key capabilities users need?
• Which of these are must-haves vs nice-to-haves?
• Are there existing products or competitors we can learn from?
• What's the smallest useful version of this?
• What screens or workflows do you envision?

Theme 6: Risks and Unknowns

• What could go wrong?
• What don't we know yet that we need to find out?
• Are there dependencies on other teams, systems, or decisions?
• What assumptions are we making that could be wrong?

Versioning

Ideation often happens in multiple rounds as thinking evolves. Each round should:

1. Increment the version in the document frontmatter (v1, v2, v3...)
2. Add a changelog entry noting what changed and why
3. Not delete previous thinking — mark superseded ideas as such rather than removing them

When Discovery receives multiple versions, it examines the delta — what changed, what was refined, what was dropped — to understand the evolution of thinking.

Output Format

Ideation produces an Ideation Document using the template at .claude/templates/ideation-document.md. The template has:

• YAML frontmatter (version, date, participants, status)
• Structured sections that map to Discovery's context questions
• A changelog for tracking version evolution

The document should be placed at docs/context/ideation-document.md in the project's specs worktree (or handed off as a standalone file if brainstorming happened outside AgentFlow).

Using This Skill Outside AgentFlow

In Claude Web or Claude Projects

Copy the "Ideation Guide" section above into your project instructions. Then start a conversation:

"I want to brainstorm a new product idea. Guide me through the ideation process and produce a structured ideation document at the end."

The AI will use the guiding questions and precision boundaries to run a structured brainstorming session. At the end, ask it to format the output using the Ideation Document template.

In ChatGPT or Other AI Tools

Copy the "Ideation Guide" section into the system prompt or custom instructions. The guiding questions and boundaries work with any capable AI.

In Workshops or Meetings

Use the 6 themes as a facilitation framework:

1. Problem Space (20 min) — What's broken?
2. Opportunity Space (15 min) — What's possible?
3. User Space (20 min) — Who benefits and how?
4. Business Context (15 min) — Why now? Why us?
5. Ideas and Features (20 min) — What could we build?
6. Risks and Unknowns (10 min) — What could go wrong?

After the workshop, consolidate notes into the Ideation Document template.

Importing Into AgentFlow

Once you have an ideation document (from any source):

1. Place it at docs/context/ideation-document.md in the project's specs worktree
2. If you have business context documents (vision statements, strategy docs, market research), place them in docs/context/ as well
3. When Discovery starts, it will automatically detect and consume these documents
4. Discovery pre-populates its 4 context questions from what it finds, then fills gaps

Integration with Discovery

How Discovery Consumes Ideation Documents

When the Discovery process skill detects documents in docs/context/:

1. Ideation documents → Extract problems, ideas, personas, business rules, constraints
2. Business context documents → Extract strategic alignment, success criteria, business drivers
3. Multiple versions → Summarize the delta: what changed, what was refined, what was dropped
4. Pre-populate Q1-Q4 → Map ideation content to Discovery's 4 mandatory questions
5. Identify gaps → Tell the user: "Based on your ideation document, here's what I already know and what I still need to ask about"

Mapping to Discovery Questions

What Discovery Adds

Even with a rich ideation document, Discovery still:

• Validates and deepens the problem understanding
• Performs technical analysis (codebase, infrastructure, integrations)
• Builds the entity model and identifies the central entity
• Makes architecture decisions (ADRs) with full technical context
• Selects the tech stack (modules) based on constraints
• Creates Linear features and milestones
• Produces the 7 mandatory Discovery deliverables

Ideation accelerates Discovery by providing rich domain context. It doesn't replace Discovery's technical analysis.

AI Guardrails

When running ideation as an AI session, enforce these boundaries:

YOU ARE facilitating an ideation session. Your role:

DO:
- Ask probing questions about problems, users, and business context
- Encourage divergent thinking — explore multiple directions
- Capture ideas without judging feasibility
- Describe user workflows and screens in detail
- Note business rules, constraints, and success metrics
- Flag assumptions and unknowns
- Use domain language (users, workflows, screens, rules)

DO NOT:
- Propose data models, database schemas, or entity relationships
- Suggest API designs, endpoints, or system architecture
- Recommend specific technologies, frameworks, or libraries
- Make infrastructure or deployment decisions
- Use technical language (tables, endpoints, schemas, microservices)
- Commit to implementation approaches

WHEN IMPLEMENTATION TOPICS ARISE:
- Capture them as "Known Constraints" if they're real constraints
  (e.g., "we use Salesforce" → constraint)
- Redirect to Discovery if they're decisions
  (e.g., "should we use GraphQL?" → "Discovery will evaluate that with full codebase context")

Common Pitfalls

1. Over-structuring too early — Ideation is messy. Don't force premature organization. Capture ideas freely, structure later.

2. Technical rabbit holes — Stakeholders may naturally discuss technology. Capture constraints, redirect decisions.

3. Skipping business context — Without the "why", Discovery can't prioritize effectively. Always capture strategic alignment.

4. Single-round thinking — Ideation often improves with reflection. Encourage multiple rounds. Each version captures evolved thinking.

5. Losing the messy bits — Superseded ideas still have value. Mark them as dropped rather than deleting — Discovery may want to understand why.

6. Too much precision too soon — Wireframe-level screen descriptions are fine. Pixel-perfect mockups are premature. Keep it at the "what the user sees and does" level.