AI Prompt Structure Standards

Beginner

Enforce consistent prompt structure across AI coding tools — role definitions, constraint formatting, output specifications, and example inclusion for reliable AI-assisted development.

File Patterns

**/CLAUDE.md**/.cursorrules**/.github/copilot-instructions.md**/.windsurfrules**/AGENTS.md

This rule applies to files matching the patterns above.

Rule Content

rule-content.md

# AI Prompt Structure Standards

## Rule
All system prompts and instruction files for AI coding tools MUST follow a consistent structure with role, constraints, output format, and examples.

## Required Structure
```markdown
## Role
[One paragraph: who the AI is and what it specializes in]

## Stack
[Bullet list of technologies with versions]

## Rules
[ALWAYS/NEVER constraints with specific patterns]

## Output Format
[How generated code should be structured]

## Examples
[At least one good/bad pair for key patterns]
```

## Constraint Formatting

### Good — Specific and Actionable
```markdown
- ALWAYS use TypeScript interfaces for component props (not type aliases)
- NEVER use `any` — use `unknown` with type narrowing instead
- ALWAYS wrap async operations in try/catch with typed error handling
- NEVER use default exports except for Next.js pages and layouts
```

### Bad — Vague and Unenforceable
```markdown
- Write good TypeScript
- Handle errors properly
- Use best practices
- Follow conventions
```

## Example Requirements
- Include at least one good/bad example pair per major pattern
- Examples should be real code, not pseudocode
- Include file path comment at the top of each example

## Length Limits
| Tool | Max Lines | Max Tokens |
|------|-----------|------------|
| CLAUDE.md | 500 lines | ~4000 tokens |
| .cursorrules | 200 lines | ~2000 tokens |
| copilot-instructions.md | 150 lines | ~1500 tokens |
| .windsurfrules | 150 lines | ~1500 tokens |

## Anti-Patterns
- "Write clean code" (unmeasurable, AI interprets differently each time)
- 50+ constraints (AI can't follow that many — prioritize top 10)
- No examples (AI guesses your patterns)
- Contradicting constraints in different sections

FAQ

Discussion

Loading comments...

# AI Prompt Structure Standards ## Rule All system prompts and instruction files for AI coding tools MUST follow a consistent structure with role, constraints, output format, and examples. ## Required Structure ```markdown ## Role [One paragraph: who the AI is and what it specializes in] ## Stack [Bullet list of technologies with versions] ## Rules [ALWAYS/NEVER constraints with specific patterns] ## Output Format [How generated code should be structured] ## Examples [At least one good/bad pair for key patterns] ``` ## Constraint Formatting ### Good — Specific and Actionable ```markdown - ALWAYS use TypeScript interfaces for component props (not type aliases) - NEVER use `any` — use `unknown` with type narrowing instead - ALWAYS wrap async operations in try/catch with typed error handling - NEVER use default exports except for Next.js pages and layouts ``` ### Bad — Vague and Unenforceable ```markdown - Write good TypeScript - Handle errors properly - Use best practices - Follow conventions ``` ## Example Requirements - Include at least one good/bad example pair per major pattern - Examples should be real code, not pseudocode - Include file path comment at the top of each example ## Length Limits | Tool | Max Lines | Max Tokens | |------|-----------|------------| | CLAUDE.md | 500 lines | ~4000 tokens | | .cursorrules | 200 lines | ~2000 tokens | | copilot-instructions.md | 150 lines | ~1500 tokens | | .windsurfrules | 150 lines | ~1500 tokens | ## Anti-Patterns - "Write clean code" (unmeasurable, AI interprets differently each time) - 50+ constraints (AI can't follow that many — prioritize top 10) - No examples (AI guesses your patterns) - Contradicting constraints in different sections