vibe-innovation

Vibe Innovation Lab

A prompt-based framework for human-AI co-creation across the full innovation lifecycle. From pre-concept exploration to validated learning. Works with any LLM (Large Language Model) that follows structured prompts.

What this is

The Vibe Innovation Framework guides you from a vague idea (or no idea at all) to a tested prototype with a clear Go, Kill, or Pivot decision. An AI acts as your thinking partner: it asks questions, challenges assumptions, structures your thinking, and generates prototype code when you are ready to build.

You do not need to write code or have technical skills to use this framework. You describe what you want in plain language. The AI does the rest.

Key terms

Term	Meaning
TRL (Technology Readiness Level)	A scale from -2 to 4 that measures how mature your innovation is. -2 means "we do not even know what to build yet." 4 means "we built it, tested it, and know it works."
ICD (Innovation Canvas Document)	A shared document that the AI fills in as you work through the phases. It accumulates all your findings, decisions, and evidence in one place.
MVP (Minimum Viable Product)	The simplest possible version of your product that lets you test whether people actually want it.

Getting started: step by step

Option A: install as Claude Code plugin (recommended)

Use this when you have a local Claude Code installation and want the framework available in any working directory.

1. In Claude Code, run /plugin marketplace add periodicpoint/vibe-innovation-lab.
2. Run /plugin install vibe-innovation@vibe-innovation-lab.
3. Invoke /vibe-innovation:innovate to start the entry diagnostic.
4. The skill loads the framework from the plugin cache, asks the diagnostic questions, and guides you through the phases.

All nine skills are namespaced under vibe-innovation: (for example /vibe-innovation:innovate-phase, /vibe-innovation:chaos).

Option B: with Claude Code in GitHub Codespaces (no install required)

Use this for one-off runs or workshops without touching a local setup.

1. Open the repository in a GitHub Codespace (green "Code" button on GitHub, then "Create codespace on main").
2. Wait until the environment finishes loading (1 to 2 minutes).
3. Open the terminal in VS Code.
4. Type claude and press Enter to start Claude Code.
5. Type /innovate and press Enter.
6. Claude loads the framework automatically, asks your starting point, team, prior knowledge, success criteria, and biggest uncertainty, then guides you through the process.

Option C: with any other LLM (ChatGPT, Copilot, Gemini, Mistral, and so on)

1. Open the repository in a GitHub Codespace (same as above).
2. Open the file .claude/docs/orchestrator.md in the repository. Click "Raw" to see the plain text. Copy everything.
3. Open your preferred LLM in a separate browser tab.
4. Paste the entire contents into a new conversation.
5. The LLM will start the entry diagnostic and guide you through the process.

What happens next (all options)

1. The Orchestrator asks about your starting point, team, prior knowledge, success criteria, and biggest uncertainty.
2. It maps your situation to a phase and assembles a session plan (which phases to run).
3. You work through the phases. The AI asks questions, you answer. The AI fills in the ICD. Each phase closes only when its output contract is met.
4. In Phase 4 (Build), the AI generates Streamlit code. You copy it into prototype/app.py, and the app updates automatically in your browser.
5. Between sessions: copy the ICD output and save it. Paste it back when you resume.

Comparison: the three options

Aspect	Plugin install (Option A)	Codespaces (Option B)	Any other LLM (Option C)
Setup	/plugin install vibe-innovation@vibe-innovation-lab	Open repo in Codespaces, run claude	Copy .claude/docs/orchestrator.md into LLM chat
Skill prefix	/vibe-innovation:innovate	/innovate	None, free-form chat
Framework loading	Automatic from plugin cache	Automatic from .claude/	Manual paste
Phase transitions	Automatic dispatch	Automatic dispatch	Say "Bitte weiter mit Phase N"
ICD management	Saved as file in working project	Saved as file in the project	Copy-paste between sessions
Prototyping (Phase 4)	AI writes to prototype/app.py of your project	AI writes to repo prototype/app.py	AI generates code, you copy it
State between sessions	Preserved in project files	Preserved in project files	You must save and re-paste the ICD
Cost	Requires Anthropic API key or Claude subscription	Anthropic key plus Codespaces minutes	Works with free LLM tiers

All paths run the same process, the same phases, and produce the same outputs.

Framework overview