AI Execution Contract

A reusable AI product thinking toolkit for turning a product idea into a structured execution contract you can use with ChatGPT, Claude, or Cursor to ship the build. Same prompts, template, and workflow in any of these tools—no lock-in.

When this repository is loaded as a skill or context source (e.g. in Cursor, Claude, or ChatGPT), the AI can guide you step-by-step through defining a project and generating an AI Execution Contract, then optionally derive UX structure or technical architecture from it.

---

Facilitator Workflow

The repository defines a guided facilitation workflow. The AI can behave like a product discovery facilitator: it asks structured questions, collects context progressively, and only generates the contract when enough information is in place. It can then help refine the contract, derive a UX structure, or suggest a technical architecture.

End-to-end process:

Idea → Facilitated questions → AI Execution Contract → (optional) UX structure → (optional) Technical architecture → Implementation

1. Idea — You have a product or feature in mind.
2. Facilitated questions — The AI asks one question at a time (project idea, problem, hypothesis, user, flow, sections, features, design readiness, assets, stack, constraints). You answer; the AI collects.
3. AI Execution Contract — The AI generates a full contract in markdown (or you use the one-shot prompt with a written brief). The contract captures problem, hypothesis, user, product structure, constraints, design system readiness, asset readiness, technical stack, and execution rules.
4. UX structure (optional) — From the contract, the AI can produce a UX structure: page/screen structure, section hierarchy, interaction logic, main flow. Product structure, not visual design.
5. Technical architecture (optional) — From the contract, the AI can suggest folder structure, component architecture, asset organization, and design system mapping.
6. Implementation — You use the contract (and optionally the UX structure or architecture) when prompting the AI to build: paste or reference the relevant sections so the model has context and rules.

---

Quick start: activate the facilitator

To use the guided facilitator in your AI tool:

1. Load this repository as context (e.g. add the repo or the workflow/ folder to your Cursor project, or paste the relevant workflow files into Claude/ChatGPT).

2. Tell the AI to use the facilitator workflow. For example:

   "Use the facilitator workflow from this repository to help me create an AI Execution Contract."

3. The AI will briefly explain the facilitator, say it will ask structured questions one at a time, and start with the first question: "What product or feature are you trying to build?"

4. Answer each question. When enough information is collected, the AI will summarize what it has, then generate the full contract. You can then refine it, generate a UX structure, or generate a technical architecture.

Alternative: one-shot generation (no facilitator)

• Open prompts/generate-contract-prompt.md, copy the short or detailed prompt, paste into your AI tool, and replace the placeholder with your project description. Copy the output into a .md file. Use prompts/refine-contract-prompt.md to refine.

---

What an AI Execution Contract is

The AI Execution Contract is an intermediate artifact between product thinking and implementation. It captures:

• Problem — What problem this product solves.
• Hypothesis — What we believe will happen if we build it.
• User — Target user, need, primary action, core flow.
• Product structure — Key sections/screens, core features, trust/validation elements.
• Constraints — Technical, compliance, performance, accessibility.
• Design system readiness — What’s ready (tokens, components) or “none yet.”
• Asset readiness — What assets exist or are placeholders.
• Technical stack — Technologies and execution rules the build must follow.

This contract helps AI-assisted tools generate better UX structures and code because they have clear context and rules. It is not just documentation—it is an execution artifact you use when prompting for implementation.

---

Design principles

This repository is built on a few principles:

• AI amplifies organization; it does not replace it. Better execution comes from better context. The contract is how you give that context.
• The AI Execution Contract is not documentation—it is an execution artifact. It is meant to be used during implementation, not only filed away.
• Design structure, design system readiness, and asset readiness are different preparation layers. The contract makes each explicit so you and the AI know what is ready and what is not.
• This system works for both solo builders and product teams. One person can run the facilitator and own the contract; teams can use it as a shared brief and align on readiness and constraints before building.

---

What’s in the repo

Item	Use
template/project-contract-template.md	Blank contract; structure for AI output.
prompts/generate-contract-prompt.md	One-shot: generate a contract from a project description (short + detailed).
prompts/refine-contract-prompt.md	One-shot: refine an existing contract (paste contract + prompt).
workflow/start-facilitator.md	Entry point: how the AI starts the facilitator (explain, set expectations, first question).
workflow/facilitator-mode.md	Behavioral instructions: guide step-by-step, ask questions, collect progressively, summarize before generating.
workflow/facilitator-questions.md	Structured question sequence (one at a time): project idea → problem → hypothesis → user → flow → sections → features → trust → design readiness → assets → stack → constraints → success → next step.
workflow/generate-contract-from-session.md	How to generate the full contract from collected session answers; template structure; assumptions vs confirmed.
workflow/refine-contract.md	Refinement workflow: analyze → propose improvements → then rewrite contract.
workflow/contract-to-ux.md	Prompt: contract → UX structure (pages/screens, section hierarchy, interaction logic, main flow).
workflow/contract-to-architecture.md	Prompt: contract → technical architecture (folder structure, components, assets, design system mapping).
workflow/benchmark-mode.md	Prompt: benchmark similar products (common UX structures, trust elements, feature sets, interaction patterns).
examples/	Full example contracts (credit card landing, finance dashboard).
docs/	Per-tool guides (ChatGPT, Claude, Cursor), PDF export, best practices.

---

Workflow (summary)

With facilitator: Idea → Facilitated questions → Contract → (optional) Refine / UX structure / Architecture → Implementation.

Without facilitator: Brief → Generate prompt → Contract → (optional) Refine → Implementation.

In both cases: use the contract (and optionally UX structure or architecture) when you prompt the AI to build. Update the contract as the project evolves.

---

Tool-specific guides

• ChatGPT
• Claude
• Cursor

Export to PDF · Best practices

---

Repository structure

ai-execution-contract/
├── README.md
├── LICENSE
├── template/
│   └── project-contract-template.md
├── prompts/
│   ├── generate-contract-prompt.md
│   └── refine-contract-prompt.md
├── workflow/
│   ├── start-facilitator.md
│   ├── facilitator-mode.md
│   ├── facilitator-questions.md
│   ├── generate-contract-from-session.md
│   ├── refine-contract.md
│   ├── contract-to-ux.md
│   ├── contract-to-architecture.md
│   └── benchmark-mode.md
├── examples/
│   ├── credit-card-landing-example.md
│   └── dashboard-example.md
└── docs/
    ├── how-to-use-with-chatgpt.md
    ├── how-to-use-with-claude.md
    ├── how-to-use-with-cursor.md
    ├── exporting-to-pdf.md
    └── best-practices.md

---

License

MIT. See LICENSE.