An AI-powered innovation methodology system that guides teams from raw idea to validated opportunity through a structured, evidence-driven process.
Stop guessing. Start validating. With evidence, not hope.
The Innovation Skills Pack is a collection of AI skills that together form a complete innovation validation workflow. Each skill is a disciplined methodology that transforms one type of input into a higher-quality output — and hands it cleanly to the next skill.
It is designed to be used with AI coding assistants (Claude Code, Antigravity IDE, OpenAI Codex CLI, Cursor).
No installation. No dependencies. Clone and start talking.
git clone https://github.com/mohanmax/Innovation-Skills.git
cd Innovation-SkillsOr download as a ZIP from the GitHub repo and extract it locally.
claudeClaude Code automatically reads CLAUDE.md as the agent orchestrator. Just describe your idea.
Open the project folder in Antigravity IDE. The AGENTS.md file is picked up automatically. Start a conversation and describe your idea.
codexThe agent reads AGENTS.md and routes your intent automatically.
- Open this folder in Cursor (
File → Open Folder) - Open AI chat (
Cmd+L/Ctrl+L) - Describe your idea or problem
Paste the contents of CLAUDE.md as your system prompt, then paste the relevant SKILL.md file for the skill you want to run.
Just describe your idea, problem, or where you are in the process:
"I have an idea for an AI tool that helps SMBs automate invoice follow-up"
→ Activates: Full Innovation Cycle (Skills 1–6)
"Just frame my idea for now, I don't need research yet"
→ Activates: Quick Framing (Skill 1 only)
"I already have research done. I want to generate hypotheses and design experiments"
→ Activates: Validation Sprint (Skills 3–6)
"I have 5 interview transcripts and I want to analyze them"
→ Activates: Interview Analysis (Skill 6 directly)
This system is built on one fundamental principle:
| What this system does | What this system does NOT do |
|---|---|
| Structure your thinking | Make business decisions for you |
| Find and evaluate evidence | Manufacture certainty |
| Design testable experiments | Execute or fake experiments |
| Interpret customer evidence | Replace customer conversations |
| Identify what you don't know | Pretend uncertainty doesn't exist |
Two human gates are non-negotiable:
- You select which hypotheses to test (after Skill 3)
- You approve the experiment before it runs (after Skill 4)
The AI informs your decisions. It does not make them.
RAW IDEA
│
▼
┌─────────────────────┐
│ 1. IDEA FRAMING │ Separates known/assumed/inferred/unknown. Captures requirements.
│ │ Identifies tensions, risks, and research questions.
└─────────────────────┘
│ → innovation-brief.md
▼
┌─────────────────────┐
│ 2. RESEARCH │ Investigates the opportunity with real, traceable evidence.
│ │ Tests assumptions. Tracks contradictions. Identifies gaps.
└─────────────────────┘
│ → research-report.md
▼
┌─────────────────────┐
│ 3. HYPOTHESIS │ Converts research into testable hypotheses.
│ │ Prioritizes transparently. AI recommends top set.
└─────────────────────┘
│ → hypothesis-portfolio.md
▼
◆ YOU SELECT which hypotheses to test ◆
│
▼
┌─────────────────────┐
│ 4. EXPERIMENT │ Designs the smallest credible experiment.
│ DESIGN │ Defines evidence, metrics, success/falsification criteria.
└─────────────────────┘
│ → experiment-design.md
▼
◆ YOU APPROVE before execution ◆
│
▼
┌─────────────────────┐
│ 5. CUSTOMER │ Designs interview protocol. Guides evidence capture.
│ INTERVIEW │ Focuses on behavior, recent events, workarounds.
└─────────────────────┘
│ → interview-study.md + interview-evidence.md
▼
┌─────────────────────┐
│ 6. INTERVIEW │ Analyzes evidence. Finds patterns, contradictions,
│ ANALYSIS │ segments. Assesses hypotheses. Extracts learning.
└─────────────────────┘
│
▼
Evidence-Backed Understanding
| # | Skill | What It Does | Output |
|---|---|---|---|
| 1 | Idea Framing | Turns a raw idea into a structured Innovation Brief | innovation-brief.md |
| 2 | Research | Investigates opportunity with real, traceable evidence | research-report.md |
| 3 | Hypothesis | Generates complete portfolio of testable hypotheses | hypothesis-portfolio.md |
| 4 | Experiment Design | Designs smallest credible experiment for selected hypotheses | experiment-design.md |
| 5 | Customer Interview | Protocol design + structured evidence capture | interview-study.md |
| 6 | Interview Analysis | Interprets evidence, finds patterns, assesses hypotheses | interview-analysis.md |
| Workflow | When to Use | Skills Invoked |
|---|---|---|
full-innovation-cycle |
Starting from a raw idea | 1 → 2 → 3 → 4 → 5 → 6 |
quick-framing |
Just frame an idea | 1 only |
research-sprint |
Frame + investigate | 1 → 2 |
validation-sprint |
Already have a brief, jump to validation | 3 → 4 → 5 → 6 |
All artifacts are saved in memory/projects/<project-slug>/:
memory/
projects/
<project-slug>/
innovation-brief.md
research-report.md
hypothesis-portfolio.md
selected-hypotheses.md
experiment-design.md
interview-study.md
interview-evidence.md
interview-analysis.md
Project slug = kebab-case from your idea name (max 40 chars). Example: ai-invoice-reconciler.
The AI helps you structure thinking, find evidence, design tests, and interpret evidence — but does not manufacture certainty.
- Research ≠ Validation
- Hypothesis ≠ Evidence
- Interview opinion ≠ Customer behavior
- Experiment design ≠ Experiment result
- Evidence ≠ Business decision
Innovation-Skills/
├── CLAUDE.md ← Root orchestrator (Claude Code / Antigravity)
├── AGENTS.md ← Root orchestrator (OpenAI Codex CLI / Cursor)
├── README.md ← This file
├── skills/
│ ├── README.md
│ ├── idea-framing/SKILL.md ← Skill 1 (v1.2)
│ ├── research/SKILL.md ← Skill 2 (v1.1)
│ ├── hypothesis/SKILL.md ← Skill 3 (v1.2)
│ ├── experiment-design/SKILL.md ← Skill 4 (v1.1)
│ ├── customer-interview/SKILL.md ← Skill 5 (v1.1)
│ └── interview-analysis/SKILL.md ← Skill 6 (v1.0)
├── workflows/
│ ├── README.md
│ ├── full-innovation-cycle.md
│ ├── quick-framing.md
│ ├── research-sprint.md
│ └── validation-sprint.md
└── memory/
└── README.md ← Project artifacts are saved here
This is an evolving methodology. Contributions, forks, and feedback are welcome.
- Found a guardrail gap? Open an issue.
- Built a new skill or workflow? Submit a PR.
- Want to adapt this for a specific domain? Fork it.
MIT — free to use, fork, and build on.