MrHaseeb
Book Call

Agentic Code Prompt Generator

Turn project type and notes into a paste-ready Cursor meta-prompt—so the assistant produces the roadmap, not this tool.

The Agentic Code Prompt Generator produces a paste-ready meta-prompt for Cursor (or any agentic IDE) that forces the AI to produce a structured implementation plan before writing code. It covers assumptions, repo layout, stack with version placeholders, conventions, phased sprints, and acceptance criteria — so you start from a plan, not vibes.

Output

Submit the form to see output here.

Advertisement

When to use this tool

  • Kicking off a new Next.js, React Native, or backend project with a clean first-pass plan.
  • Onboarding a contractor to an unfamiliar stack by having the agent produce a conventions doc upfront.
  • Forcing the AI to ask the right clarifying question instead of inventing answers.
  • Turning a vague client brief into a scoped sprint plan a team can actually estimate.

How it works

  1. Pick the project type closest to your build (landing page, SaaS, mobile app, etc.).
  2. Add context — constraints, team size, integrations, deadlines, anything the plan should respect.
  3. Optionally name the UI library / stack you are already committed to.
  4. Paste the generated meta-prompt into Cursor and let the assistant return the plan before any code.

Example output

Sample only — your generated output will reflect your specific inputs.

**Meta-prompt (paste into Cursor):**

You are a technical lead. The user is building a **marketing landing page**. Your job in this chat is **not** to write code yet—produce a **structured implementation plan** with: assumptions, repo layout (tree), stack with **version placeholders** (never invent semver), conventions, 4–6 phased sprints with deliverables, and testing/acceptance criteria. Ask at most one clarifying question if scope is ambiguous.

_(Offline preview — Generate for a tailored meta-prompt.)_

Tips for best results

  • Keep 'Context & goals' specific — deadlines and constraints beat generic goals.
  • Ask the agent to list open questions at the end of the plan before writing code.
  • Re-run the plan every time scope changes, not just at kickoff.

Frequently asked questions

Does this work with models other than Claude or GPT?

Yes. The prompt structure is model-agnostic and has been tested with Claude 3.5/4, GPT-4o/5, and Gemini 2.5.

Why does the prompt forbid specific version numbers?

LLMs routinely invent semver. The prompt uses version placeholders so you pin the real versions from package managers, not from memory.

Can I use this for refactors instead of greenfield projects?

Yes. Add the existing repo's structure and conventions into the 'Context & goals' field and the agent will produce a refactor plan that respects them.

What makes this different from a normal prompt?

Normal prompts jump to code. This meta-prompt forces the AI into a planning role first, so you get architecture before implementation.