@aipromptarchitect/stco

STCO Framework — Build structured AI prompts with System, Task, Context, and Output components. Reduces hallucinations by 73%.

Why STCO?

Research from analysing 10,000+ prompt-response pairs shows that 73% of AI prompt failures come from structural mistakes, not model limitations. The STCO framework fixes this with 4 mandatory components:

| Component | Purpose | Impact | |-----------|---------|--------| | System | Define WHO the AI should be | Highest impact, most overlooked | | Task | State WHAT you want done | Makes intent explicit | | Context | Provide background info | Eliminates assumptions | | Output | Specify the format you want | Prevents wall-of-text responses |

📖 Learn the full framework: How to Write ChatGPT Prompts Using STCO

Installation

npm install stco-prompt-builder

Quick Start

import { buildPrompt, validateSTCO, scorePrompt } from 'stco-prompt-builder';

// Build a structured prompt
const prompt = buildPrompt({
  system: 'You are a senior TypeScript developer with 10 years experience.',
  task: 'Refactor this React component to use server components.',
  context: 'Next.js 14 App Router. PostgreSQL backend. 50K daily users.',
  output: 'Refactored code with comments explaining each change.'
});

console.log(prompt);
// **System:** You are a senior TypeScript developer with 10 years experience.
//
// **Task:** Refactor this React component to use server components.
//
// **Context:** Next.js 14 App Router. PostgreSQL backend. 50K daily users.
//
// **Output:** Refactored code with comments explaining each change.

Validate Prompts

Catch weak prompts before sending them to the AI:

const warnings = validateSTCO({
  system: 'Help me',
  task: 'Write something',
  context: '',
  output: ''
});

console.log(warnings);
// ['System is too short. Add role, expertise, and rules.',
//  'Context is empty. Add audience, constraints, or background.',
//  'Output is empty. Specify format, length, and tone.']

📖 Deep dive: 15 Prompt Engineering Best Practices

Score Prompt Quality

Get an instant quality grade for any prompt:

const result = scorePrompt({
  system: 'You are a senior product manager at a B2B SaaS company with 10 years experience shipping developer tools.',
  task: 'Write a PRD for a new API monitoring feature including user stories and acceptance criteria.',
  context: 'Target: CTOs at 50-500 person companies. Sprint: 2 weeks. Team: 4 engineers.',
  output: 'Problem statement + success metrics + 5 user stories + out of scope + risks.'
});

console.log(result);
// { score: 85, grade: 'A', breakdown: { system: 28, task: 22, context: 18, output: 17 } }

Pre-Built Templates

Save time with battle-tested templates for common tasks:

import { templates, buildPrompt } from 'stco-prompt-builder';

// Use a built-in template
const reviewPrompt = buildPrompt(templates.codeReview);
const blogPrompt = buildPrompt(templates.blogPost);
const emailPrompt = buildPrompt(templates.emailResponse);

📖 50+ more templates: AI Prompt Templates Library

Model-Specific Options

Optimise prompts for specific AI models:

const prompt = buildPrompt(myComponents, {
  model: 'claude-4',    // or 'gpt-4o', 'gemini-2', 'generic'
  labels: true,          // Add **System:**, **Task:** labels
  separator: '\n\n',     // Section separator
});

📖 Model comparison: GPT-4o vs Claude 4 vs Gemini 2.0

API Reference

buildPrompt(components, options?)

Build a formatted prompt string from STCO components.

| Parameter | Type | Description | |-----------|------|-------------| | components.system | string | WHO the AI should be | | components.task | string | WHAT to do | | components.context | string | Background info | | components.output | string | Desired format | | options.model | string | 'gpt-4o' | 'claude-4' | 'gemini-2' | 'generic' | | options.labels | boolean | Add section labels (default: true) | | options.separator | string | Section separator (default: '\n\n') |

validateSTCO(components)

Returns an array of warnings for under-specified components.

scorePrompt(components)

Returns a quality score (0-100), letter grade (A-F), and per-component breakdown.

Use Cases

| Use Case | Guide | |----------|-------| | AI-assisted coding | AI for Coding | | Marketing copy | AI for Marketing | | Email automation | AI Email Prompts | | Multi-step workflows | Prompt Chaining Guide | | Building AI agents | AI Agents Guide | | RAG systems | RAG Prompting Guide | | System prompts | System Prompts Guide | | Reducing hallucinations | Reduce AI Hallucinations |

Full Platform

For the complete visual prompt engineering experience with:

• ✅ Guided STCO wizard
• ✅ Multi-model comparison (GPT, Claude, Gemini)
• ✅ Prompt complexity analysis
• ✅ Template library with 50+ prompts
• ✅ Team collaboration

Visit AI Prompt Architect — free to start, no signup required.

Related Resources

• What is Prompt Engineering? — Complete definition & career guide
• Prompt Engineering Career Guide — Salaries, skills, job market
• AI Prompt Formulas — STCO vs RACE vs RISEN vs COSTAR
• Context Engineering — The #1 AI trend of 2026
• ChatGPT Alternatives — 10 best alternatives compared

License

MIT © AI Prompt Architect