Prompting Starter Kit for Builders (From Official Guides)

Most people search for 'prompt examples' and copy random templates, but results stay inconsistent because the prompt lacks clear constraints, output format, and acceptance criteria.

Official guides agree on one core pattern: define role, objective, context, constraints, and expected output shape before asking for content.

• Common mistake #1: broad requests without measurable success criteria.
• Common mistake #2: mixing optional ideas with mandatory requirements.
• Common mistake #3: no output contract, so responses become verbose and hard to reuse.

Use this structure for almost every technical prompt. It is easy to learn and works across OpenAI, Claude, and Gemini-style models.

• Role: who the model should act as (principal engineer, PM, reviewer).
• Goal: exact outcome and definition of done.
• Context: product constraints, user type, stack, and scope.
• Rules: non-negotiables, limits, and things that must not change.
• Output contract: sections, format, and acceptance checks.

Start from this template and edit only the INPUT block. Keep the contract and constraints stable to get repeatable outputs.

Role:
You are a senior product engineer.

Goal:
Produce a build-ready plan I can execute today.

Constraints:
- No generic advice.
- Map recommendations to files, endpoints, tables, or tests.
- Keep scope realistic for an MVP.
- Language: Write the entire output in Italian, but keep technical terms, code snippets, and folder names in English.

Output contract:
1) Recommended stack with reasons
2) Architecture + data model
3) Implementation phases with acceptance criteria
4) Testing strategy (unit/integration/e2e)
5) Risks + launch checklist

### INPUT ###
Idea: <your idea>
Required constraints (MUST include):
- <constraint 1>
- <constraint 2>
### END INPUT ###

These prompts are optimized for practical project execution and reduce rework during implementation.

• PRD prompt: 'Turn this idea into a lean PRD with user stories, success metrics, and non-goals.'
• Architecture prompt: 'Given this PRD, propose a production-ready architecture with schema + API contracts.'
• Execution prompt: 'Convert architecture into 2-week phases with acceptance criteria and test plan.'

A short QA pass catches most low-quality responses before coding starts.

• Are requirements explicit and measurable?
• Is output structured in reusable sections?
• Are assumptions listed instead of hidden?
• Is there at least one schema/API/test artifact?
• Can another developer execute it without extra clarification?

1. OpenAI Prompting Guide(developers.openai.com)
2. OpenAI Prompt Engineering Guide(developers.openai.com)
3. Anthropic Prompt Engineering Overview(docs.anthropic.com)
4. Anthropic XML Tags Best Practice(docs.anthropic.com)
5. Google Vertex AI Prompt Design Strategies(cloud.google.com)
6. Google Prompt Engineering Overview(cloud.google.com)
7. Microsoft Prompt Engineering (Azure OpenAI)(learn.microsoft.com)