ai-prompts

name	ai-prompts
description	AI prompt library covering system prompts, few-shot templates, structured output schemas, and prompt engineering patterns for production LLM applications
layer	domain
category	ai
triggers	["system prompt","few-shot","prompt template","structured output","prompt engineering","LLM prompt","chain of thought","prompt library"]
inputs	["Task description requiring prompt design","Existing prompts needing optimization","Structured output format requirements","Few-shot example construction needs"]
outputs	["Production-ready system prompts","Few-shot template sets with examples","Structured output schemas (JSON, XML)","Prompt evaluation and improvement suggestions"]
linksTo	["prompt-engineering","ai-agents","rag"]
linkedFrom	["ai-agents","prompt-engineering"]
preferredNextSkills	["prompt-engineering","ai-agents"]
fallbackSkills	["rag"]
riskLevel	low
memoryReadPolicy	selective
memoryWritePolicy	none
sideEffects	[]

AI Prompt Library

Purpose

Provide ready-to-use prompt patterns, system prompt templates, few-shot designs, and structured output schemas for building reliable LLM-powered applications. Focuses on deterministic, testable prompts that produce consistent results across models.

Key Patterns

System Prompt Architecture

Role-Task-Format-Constraints (RTFC) framework:

# System Prompt Template

## Role
You are a {role_description} with expertise in {domain}.

## Task
{clear_task_description}

## Input Format
You will receive: {input_description}

## Output Format
Respond with: {output_specification}

## Constraints
- {constraint_1}
- {constraint_2}
- If uncertain, {fallback_behavior}

## Examples
{few_shot_examples}

Production system prompt example:

const CODE_REVIEWER_PROMPT = `
You are a senior code reviewer specializing in TypeScript and React.

## Task
Review the provided code diff and identify:
1. Bugs or logic errors
2. Security vulnerabilities
3. Performance issues
4. Style/readability improvements

## Output Format
Respond with a JSON array of findings:
\`\`\`json
[{
  "severity": "critical" | "warning" | "info",
  "line": number,
  "category": "bug" | "security" | "performance" | "style",
  "description": "What the issue is",
  "suggestion": "How to fix it"
}]
\`\`\`

## Constraints
- Only report genuine issues, not style preferences
- Maximum 10 findings, prioritize by severity
- If no issues found, return an empty array []
- Never suggest changes that alter behavior without flagging as "breaking"
`;

Few-Shot Templates

Consistent formatting drives better results:

function buildFewShotPrompt(
  task: string,
  examples: { input: string; output: string }[],
  query: string
): string {
  const exampleBlock = examples
    .map((ex, i) => [
      `### Example ${i + }`,
      ,
      ,
    ].())
    .();

   ;
}


 prompt = (
  ,
  [
    { : , :  },
    { : , :  },
    { : , :  },
  ],
  userReview
);