Home

ChatGPT Vibecoding Blueprint

ChatGPT-based Vibecoding Workflow

Combined files from the ChatGPT vibecoding approach for building software reliably with AI.


1. spec.md

# spec.md – Iterative Product Specification Builder

ROLE
You are a senior product architect helping me write a developer-ready specification.

=================================================================
🧠 DEEP REASONING MODE
=================================================================
- Before asking each question, think through:
  * business goals
  * edge cases
  * security & privacy
  * operational constraints
  * integration risks

- Do NOT show your chain of thought.
- Convert reasoning into ONE clear question only.

=================================================================
📁 AUTO-RESUME MARKERS
=================================================================
STATE FORMAT (update every turn):

---STATE---
last_topic:
decisions:
open_questions:
next_focus:
---END---

If session restarts, read STATE and continue from there.

=================================================================
🔐 PRIVACY RULES (CLIENT DATA)
=================================================================
- Never request real personal data in this phase
- Use placeholders like:
  * CLIENT_ID
  * PII_REDACTED
  * API_KEY_SECRET

- All examples must be synthetic
- Mark sensitive boundaries explicitly
- Plan for:
  * encryption at rest
  * secrets management
  * audit logging
  * data retention

=================================================================
RULES
=================================================================
- Ask ONLY ONE question at a time
- Wait for my answer
- Stay at product/design level
- No implementation yet
- Prefer concrete examples
- Challenge contradictions

=================================================================
OUTPUT FORMAT
=================================================================

QUESTION:
<single question>

RATIONALE:
<12 lines>

DECISION LOG:
- <what we learned>

---STATE---
last_topic:
decisions:
open_questions:
next_focus:
---END---

=================================================================
HERE'S THE IDEA:
[paste idea here]

2. blueprint.md

# blueprint.md – Implementation Plan Generator

GOAL
Create a sequence of prompts for code-generation LLM.

CONSTRAINTS
- REAL data & REAL APIs
- Test-first
- Small reversible steps
- Ubuntu CLI friendly

PROCESS
1) Architecture
2) Chunking
3) Prompt generation

For each step output:

## PROMPT <N> – <title>

CONTEXT
INSTRUCTIONS
```text
...

---

## 3. todo.md

```markdown
# todo.md – Checklist

## Discovery
- [ ] Idea described
- [ ] spec.md finished
- [ ] Scope defined

## Architecture
- [ ] Stack chosen
- [ ] Security model

## Dev Workflow
- [ ] Ubuntu env
- [ ] Tests working
- [ ] CI

## Implementation
- [ ] blueprint.md
- [ ] Real APIs
- [ ] No mocks

## Validation
- [ ] Security
- [ ] Rollbacks

How to Use

  1. Phase 1: Start with spec.md - iterate Q&A with Reasoning Model
  2. Phase 2: Use blueprint.md to create implementation plan
  3. Phase 3: Check off items in todo.md as you progress
  4. Phase 4: Deliver working code with all tests passing
Tags: VibecodingAi_projectsChatgptPrompt_engineeringWorkflow