| name | cursor-project-starter |
| description | Generates a complete Cursor AI IDE project starter kit — 7 documents — from a brainstorming conversation with the user. Use this skill whenever the user wants to start a new project, says "let's start a new project", "I have a project idea", "help me plan a new app/system/tool", "set up my Cursor project", or wants to brainstorm and immediately get working project files. Also trigger when the user says things like "I want to build X from scratch" or "help me plan out X before I start coding". Always use this skill proactively when a new project is being initiated — don't wait for the user to ask for documentation explicitly.
|
Cursor Project Starter Kit
Generates 7 structured project documents through a brainstorming interview, ready to drop
into Cursor (or any AI IDE) so the AI understands the full project from the start.
The 7 Documents This Skill Produces
| File | Purpose |
|---|
PRD.md | Product Requirements — features, goals, metrics, risks |
technical-spec.md | Architecture, tech stack, functions, API details |
app-flow.md | User journeys, system states, edge cases |
schema.md | Database/data structure — all tables/collections + columns |
implementation-plan.md | Phased build checklist with tasks and checkboxes |
README.md | Project index — explains every file, how to use in Cursor |
.cursorrules | Cursor AI rules — code style, naming, patterns, anti-patterns |
Step 1 — Brainstorm Interview
Before writing any documents, conduct a structured interview. Ask ALL of these questions in a
single conversational message (not one at a time). Group them naturally.
Core Questions
What is the project?
- What does it do in one sentence?
- Who are the users (primary + secondary)?
- What problem does it solve?
Tech Stack:
- What language/runtime? (e.g., Python, Node.js, Google Apps Script, React)
- What database or storage? (e.g., PostgreSQL, Firestore, Google Sheets, MongoDB)
- What external APIs or services? (e.g., OpenAI, Stripe, WhatsApp, Google Maps)
- Any hosting/deployment target? (e.g., Vercel, GCP, AWS, serverless)
Features:
- What are the 3–5 core features for MVP?
- What is explicitly OUT OF SCOPE for now?
- Any "nice to have" features?
Data:
- What are the main data entities? (e.g., users, orders, products, sessions)
- What fields does each entity need?
Flows:
- Walk through the main user journey step by step
- Are there admin/owner flows separate from customer flows?
- What are 2–3 key edge cases or failure scenarios?
Constraints:
- Timeline or deadline?
- Cost constraints (free tier only? budget ok?)
- Any security or compliance requirements?
Step 2 — Fill Gaps & Confirm
After the user answers:
- Infer anything not explicitly stated (e.g., if they say "React app", infer Vite/Next.js as likely choice)
- Ask for clarification ONLY on blockers (missing DB choice, unclear primary user)
- Summarize your understanding in 3–5 bullet points
- Ask: "Does this match your vision? Ready to generate the files?"
Step 3 — Generate All 7 Documents
Generate all files in sequence. Write each one completely before moving to the next.
Use the templates in templates/ as structural guides — replace all placeholder content
with the user's actual project details.
Generation order:
PRD.md — establish requirements first (everything else references this)technical-spec.md — architecture derived from PRDschema.md — data model derived from tech specapp-flow.md — user journeys referencing the features in PRDimplementation-plan.md — phased tasks referencing all above.cursorrules — rules derived from tech stack and patterns established aboveREADME.md — index file written last (references all others)
Quality checks per document:
- No placeholder text like
[FILL IN] or YOUR_VALUE left in final output
- Tech stack must be consistent across ALL documents
- Function names in
.cursorrules must match function names in technical-spec.md
- Schema column names in
schema.md must match references in technical-spec.md
- All phase tasks in
implementation-plan.md should reference the correct companion doc
Step 4 — Package & Deliver
Save all 7 files to /mnt/user-data/outputs/ in a flat structure (no subfolder needed).
Present them to the user with present_files.
Tell the user:
"Drop all these files into your Cursor project root folder. Cursor will auto-detect
.cursorrules. Then start prompting with references like:
'Read PRD.md and technical-spec.md, then write the [function name] function'"
Document Templates
See templates/ folder for the structural template of each document.
Read the relevant template before generating that document.
templates/PRD-template.mdtemplates/technical-spec-template.mdtemplates/app-flow-template.mdtemplates/schema-template.mdtemplates/implementation-plan-template.mdtemplates/cursorrules-template.mdtemplates/README-template.md
Key Principles
- Project-specific, not generic — every document must feel written FOR this project, not copy-pasted
- Cross-referenced — documents must refer to each other (e.g., implementation-plan references PRD phases)
- Cursor-optimized — README must include ready-to-use Cursor prompt examples specific to the project
- No server assumption — if user didn't mention a server, default to serverless/managed options
- Indian context awareness — if project targets India, default to INR, Indian phone format (+91), UPI/Razorpay, Aadhaar-aware considerations where relevant
