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