nbsp1221/idea-validation-autopilot

Idea Validation Autopilot

Turn a rough idea into an evidence-backed build decision in one run.

Overview

This skill is a single orchestrator for:

1. idea clarification
2. market and competitor research
3. MVP scope definition
4. go/no-go style decision memo

Default behavior favors action over over-analysis:

• ask as few questions as possible
• run parallel research
• output a build-ready decision packet

When to Use

Use this skill when:

• user says they have many ideas but cannot decide efficiently
• user wants "startup-like" process without paying for SaaS tools
• user wants AI to drive research and synthesis, not just brainstorm
• user asks for market validation + MVP boundaries + next execution steps

Do not use this skill when:

• user already has validated requirements and only wants implementation planning
• user wants only code generation with no discovery work

Operating Defaults

If user context is missing, proceed with defaults instead of blocking:

• Goal priority: speed-to-learning > polish
• Budget assumption: near-zero external spend
• Team assumption: solo builder or very small team
• Timebox assumption: one focused discovery cycle

Only ask questions when missing data would invalidate the result (for example: unclear target user or regulated domain).

Workflow

Copy this checklist and track progress:

Progress
- [ ] Step 1: Normalize idea into problem hypothesis
- [ ] Step 2: Run 4 parallel research tracks
- [ ] Step 3: Grade evidence quality and resolve contradictions
- [ ] Step 4: Produce decision scorecard and verdict
- [ ] Step 5: Define MVP scope and exclusions
- [ ] Step 6: Define first experiments and stop rules
- [ ] Step 7: Deliver final report using template

Step 1: Normalize the idea

Convert raw idea into this structure:

• target user
• painful job-to-be-done
• current workaround
• why-now trigger
• value promise in one sentence

If unclear, propose your best assumption and mark it explicitly.

Step 2: Run 4 parallel research tracks

Dispatch four independent subagents (or equivalent parallel workers).

1. User/Problem Research

• Find who feels the pain and how urgently.
• Capture behavioral evidence, not just opinions.

2. Market/Competitor Research

• Map direct/adjacent alternatives, pricing, positioning, switching cost.
• Identify market gap with realistic differentiation.

3. Business Model/Risk Research

• Estimate willingness-to-pay signals, acquisition path, and major risks.
• Flag legal/compliance/data-access blockers early.

4. MVP/Technical Feasibility Research

• Define thinnest viable product delivering the core job.
• Identify build constraints, integration risks, and timeline risk.

Step 3: Grade evidence quality

Use evidence tiers:

• Tier A: behavioral or monetary signal (payment, waitlist intent with commitment, repeated real usage)
• Tier B: strong secondary evidence (credible reports, robust competitor/user data)
• Tier C: weak signal (opinions, generic trend articles, unsupported claims)

Rules:

• critical claims need at least two independent sources
• if evidence is weak, lower confidence regardless of narrative quality

Step 4: Score and decide

Score 0-100 using weighted dimensions:

| Dimension | Weight | | --- | ---: | | Problem severity and frequency | 25 | | Distribution reachability | 20 | | Willingness-to-pay potential | 20 | | MVP speed/feasibility | 20 | | Strategic differentiation | 15 |

Scoring rules (fixed):

• each dimension score is 0..100
• weighted_i = score_i * weight_i / 100
• total_score = round(sum(weighted_i), 1)
• map verdict from total_score using the bands below

Verdict bands:

• 80-100: Build now
• 60-79: Validate-first (run targeted tests before building)
• 40-59: Pivot
• <40: Drop

Step 5: Define MVP scope

Use strict scope slicing:

• Must: smallest set proving core value
• Should: useful but deferrable
• Won't (now): explicitly excluded features

Output a 2-week implementation target:

• week 1: build core flow
• week 2: launch to first users and collect signals

Step 6: Define experiments and stop rules

For top risks, define:

• experiment
• pass threshold
• fail threshold
• next action if pass/fail

Keep experiments cheap and fast. Favor reversible steps.

Step 7: Deliver final report

Use assets/final-report-template.md.

Output path rules:

• if reports/ does not exist, create it first (mkdir -p reports)
• write report to reports/YYYY-MM-DD-<idea-slug>-idea-validation.md

Required output qualities:

• explicit assumptions table
• explicit unknowns
• citations and dated evidence
• final recommendation plus next 7-day action plan

Common Failure Modes

1. Over-research without decisions

• Fix: enforce scorecard and verdict section every run.

2. Generic competitor list with no switching analysis

• Fix: include why users switch or stay.

3. MVP too large

• Fix: require "what can be deleted" before finalizing scope.

4. False confidence from weak sources

• Fix: downgrade to Tier C and force validation-first verdict.

Quick Command Patterns

Adapt to available tools:

• web search + fetch for sources
• repository/API lookup for existing solutions
• parallel subagents for independent tracks
• markdown report output in project reports/

If one tool is unavailable, continue with the best fallback and document the limitation in assumptions.