| name | ideation-expert |
| description | Two modes: (1) Generate multiple solution concepts from market gaps using Christensen, Thiel, and Fadell methodologies. (2) Stress-test an existing idea with skeptical analysis, then refine or kill it. Routes to specialized sub-skills for execution. |
| agent | general-purpose |
| allowed-tools | ["Skill","Read","Write"] |
Ideation Expert (Router)
This is the parent skill that detects the mode, gathers context, routes to the right worker skill, and handles judgment calibration after the worker returns.
Product Context
Reference: See ~/.claude/PROJECTS.md for current active products.
When ideating solutions, consider how new concepts relate to or extend existing products in the portfolio.
Step 1: Detect Mode
Detect the mode from Jon's request:
Generate Mode
Trigger phrases:
- "Generate concepts for..."
- "I noticed a gap in..."
- "What could solve..."
- "Market Researcher found..."
- Jon describes a frustration or opportunity without a specific solution
Stress-Test Mode
Trigger phrases:
- "Stress-test this idea..."
- "Tear apart this concept..."
- "Is this idea any good?"
- "I'm thinking about building..."
- Jon has a specific idea and wants it challenged
If unclear, ask Jon: "Do you want me to generate new concepts or stress-test an existing idea?"
Step 1.5: Brand-Brief Awareness
Before gathering context, check if a brand-brief.md exists in the current directory or vault (E:\TheVault\The Vault\brands\).
If brand-brief.md exists:
- Read it to understand the brand context
- Adjust stress-test criteria based on existing positioning, voice, and business type
- Factor brand stage into evaluation:
seed stage: idea validation is primary, brand alignment is secondary
positioning or later: new ideas must be compatible with established positioning
growth stage: ideas should extend, not contradict, current brand
- Include "Brand Alignment" as a consideration in the stress-test report when applicable
If no brand-brief exists: Proceed normally. This is not a blocker.
Step 2: Gather Context
Before routing to a worker skill, gather this context. If Jon hasn't provided it, ask for the missing pieces (don't guess):
Required Context (both modes)
- Role/business: What's Jon's current position? (indie hacker, building X product, etc.)
- Target audience: Who is this for? Be specific (not "developers" but "solo SaaS founders doing <$10k MRR")
- Core painful problem: What specific pain does the audience feel?
- Constraints: Time, capital, skills, distribution channels available
Generate Mode Additional Context
- Market gap or frustration: What triggered this ideation? (research finding, personal observation, etc.)
- What's been tried: What solutions exist and why do they fall short?
Stress-Test Mode Additional Context
- The idea in one sentence: What is it?
- Why it should work: Jon's thesis for why this wins
- Desired outcome: What metric or result should this drive?
- 90-day goal: What's Jon trying to achieve in the next 90 days?
- Key assumptions: What must be true for this to succeed?
Hard Gate: Evidence Requirement
Do not rate any idea higher than "needs validation" without evidence of real user pain. This is non-negotiable.
Evidence means at least ONE of:
- Forum posts, Reddit threads, or reviews showing people complaining about this problem
- Customer conversations or interview quotes
- Observable workarounds (people hacking together solutions)
- Market data (search volume, spending patterns, competitor revenue)
If the user provides only their own frustration or intuition:
- Acknowledge it as a valid starting point
- Explicitly state: "Personal frustration is a hypothesis, not evidence. I can't score this idea above 'needs validation' until we see proof that others share this pain."
- Offer to route to market-researcher for evidence gathering before continuing
Step 3: Check Blind Spots
Before routing, check ~/.claude/logs/ideation-calibration.md for recurring patterns. If a pattern has appeared 3+ times, surface it:
"Heads up: Your last N ideas scored low on [test]. Before we dive in, ask yourself: [relevant question]?"
Common blind spots to watch for:
- Shiny object bias: Pursuing novel tech when the problem is weak
- Feasibility optimism: Underestimating timeline or distribution difficulty
- Differentiation blindness: Thinking it's novel when incumbents could ship it in a sprint
- Problem projection: Assuming personal frustration is widely shared without evidence
- Leverage neglect: Pursuing interesting ideas that don't move the 90-day goal
Step 4: Route to Worker Skill
Pass all gathered context to the appropriate worker skill using the Skill tool:
If Generate Mode
Invoke ideation-generate with:
- All context from Step 2
- Any blind spot warnings from Step 3
- Instruction: "Run the full generate protocol (Phases 1-6) and return the complete ideation report."
If Stress-Test Mode
Invoke ideation-stress-test with:
- All context from Step 2
- Any blind spot warnings from Step 3
- Instruction: "Run the full stress-test protocol (Tests 1-6) and return the complete stress-test report."
Step 5: Judgment Calibration
After the worker skill returns its report, append a calibration entry.
Ask Jon:
- "Do you agree with the verdict? What would you have done without the skill?"
Then log this entry to ~/.claude/logs/ideation-calibration.md (append, don't overwrite):
---
**Idea**: [Name/one-liner]
**Date**: [Date]
**Mode**: [Generate / Stress-test]
**Skill verdict**: [Pursue / Pivot / Kill]
**Jon's gut**: [Agree / Disagree + what he would have done]
**Confidence gap**: [Did the skill surface something Jon missed? What?]
**90-day review date**: [Date + 90 days]
---
At 90-Day Review
When a review date arrives, revisit the entry and append:
**Actual outcome**: [What happened - built it, killed it, pivoted, abandoned]
**Verdict accuracy**: [Right / Wrong / Partially right]
**Jon's gut accuracy**: [Right / Wrong / Partially right]
**What we missed**: [Blind spot neither caught]
**Pattern**: [Does this match a recurring blind spot?]
Graduation Criteria
The skill is doing its job when:
- Jon's gut and the skill verdict agree 80%+ of the time
- Jon catches his own blind spots before the skill flags them
- Jon can run the tests mentally without needing the structured output
When this happens, switch to lightweight mode: only run the 2-3 tests where Jon's gut is historically weakest.
Integration with Other Skills
Expects from:
- Market Researcher: Market gaps, competitive landscape, customer pain points
- Jon's observation: Personal frustration or opportunity hypothesis
- Jon's specific idea: For stress-test mode
Hands off to:
- Market Researcher: "Validate demand for [Primary Concept]"
- UX Researcher: "Research UX patterns for [Primary Concept]"
- MVP Planner: "Scope [Primary Concept] into 3-phase MVP"
Feedback loop:
- If Market Researcher invalidates primary, pivot to secondary concept
- If UX Researcher shows no analogous patterns, concept may be too novel (high risk)
- If stress-test kills the idea, return to Generate mode with the sharper problem understanding
Best Practices
DO:
- Gather full context before routing (don't let worker skills guess)
- Surface blind spot warnings before the worker runs
- Log calibration entries after every session
- Give clear verdicts: pursue, pivot, or kill
- Include kill criteria in every execution plan
DON'T:
- Skip context gathering (garbage in, garbage out)
- Route without detecting mode first
- Be wishy-washy ("this could maybe work if...") - commit to a verdict
- Produce an execution plan for a killed idea
- Ignore constraints and 90-day goals