Build with AI Flow

Run the Cursor-first development loop from planning to deploy during live sessions.

Build with AI Flow

Use this run sheet while building. It keeps you in a consistent Cursor-guided loop: plan, prompt, implement, verify.


Quick Run Sheet

  1. Frame the task — Restate the user story or bug in one sentence
  2. Prep Cursor — Open the repo, highlight key files, feed the goal and constraints
  3. Generate plan — Ask for a minimal change list; confirm or trim
  4. Apply changes — Let Cursor write files, review diff together, iterate fast
  5. Verify — Run local command or deployment check; log issues in Troubleshooting

Repeat until the user story is complete and merged.


1. Kickoff and Planning

  • Pull from main and run the project. Confirm you're on the latest commit.
  • Review the worksheet outputs from Planning Worksheet to anchor scope.
  • In Cursor chat, paste the current objective and any guardrails (tech stack, UX must-haves, data constraints).

Prompt starter:

We are continuing the [Project Name] app. Current status: [brief summary].
Objective: [single user story].
Constraints: [bullets].
Return: a short implementation plan with files to touch, pre-checks, and verifications.

2. Structure the Work

  • Use the AI plan as a checklist; keep only the steps you truly need.
  • Assign roles if working in a group: one person drives Cursor, one tests, others check docs or schema.
  • Capture new tasks or risks in your worksheet backlog so nothing is forgotten.

Checkpoint: Plan fits on one screen, every step maps to a repo change or validation.


3. Pair with Cursor

  • For UI or schema updates, highlight relevant files before prompting to give Cursor context.
  • Keep prompts tight: what changed, what to create, expected output format.
  • Let Cursor write, then immediately scan the diff for misalignments. Reject and re-prompt quickly if it strays.

When Cursor guesses wrong:

  1. Explain the mismatch ("We need Supabase Row Level Security, not fake auth.")
  2. Paste the correct snippet or docs link
  3. Re-run with the clarified constraint

4. Test and Iterate

After each meaningful change, run the smallest verification possible:

  • npm run dev or vercel dev for frontend
  • supabase db diff / db push for schema
  • Unit or integration tests if available

Note any failing command, then jump to the matching section in Troubleshooting.

Capture resolved issues so you build a troubleshooting habit over time.


5. Merge and Deploy

  • Once the story works locally, create a focused commit:
git commit -m "feat: add RSVP form validation"
  • Push and open a pull request; review line-by-line.
  • Merge to main when approved. Let Vercel auto-deploy or run vercel --prod.
  • Celebrate the win, then queue the next objective.

Sample Prompts Library

Generate Supabase Migration

You are designing a PostgreSQL migration for Supabase.
Task: add a `restaurant_preferences` table tied to `events`.
Include indexes for lookups by `event_id`, set defaults, and enable RLS (policies later).
Return only SQL.

Update a UI Component

We added Supabase restaurant preferences. Update `components/RestaurantForm.tsx`
to display existing preferences and add a toggle for `requires_accessibility`.
Keep styling consistent with Tailwind classes already in the file.

Keep adding winning prompts here so each session benefits from the last.


Timeboxing Tips

PhaseTime
Plan + prep prompts15 minutes
Implement with Cursor25 minutes
Verify, document, push10 minutes

Adjust the timer but keep the rhythm. Flow comes from short feedback loops.


Key Principles

  • Plan with intention — Know what you're building before you prompt
  • Prompt with clarity — Be specific about constraints and expected output
  • Verify relentlessly — Don't trust; confirm

Cursor becomes the accelerator, not the driver. You stay in control.


Ready to debug? See Troubleshooting & Logs.

Need CLI commands? Check CLI Services.