Prompt Best Practices

Get the most out of Nexus Prompts with these proven patterns and recommendations.

Organizing Your Prompts

Create a Prompt Hierarchy

Build prompts that progressively specialize:

Level 0: Base Prompts
├── typescript-fundamentals
├── testing-fundamentals
└── devops-fundamentals

Level 1: Technology Stacks
├── react-typescript (extends typescript-fundamentals)
├── node-typescript (extends typescript-fundamentals)
└── fullstack-typescript (extends both)

Level 2: Project-Specific
├── my-saas-frontend (extends react-typescript)
└── my-saas-backend (extends node-typescript)

One Prompt Per Concern

Rather than one large prompt, use multiple focused prompts:

# Multiple active prompts for a testing session
nexus prompt activate react-typescript-expert
nexus prompt activate testing-qa-engineer

# Prompts are merged: React knowledge + Testing expertise

Priority Ordering

When multiple prompts are active, earlier prompts take precedence for conflicts:

# Priority 1 (highest)
nexus prompt activate react-typescript-expert --priority 1

# Priority 2
nexus prompt activate testing-qa-engineer --priority 2

Writing Effective System Prompts

Be Specific About Constraints

Instead of:

Be careful with security.

Write:

Security Requirements:
- Always sanitize user inputs before database queries
- Use parameterized queries, never string concatenation
- Validate file uploads: check MIME types, limit size to 5MB
- Hash passwords with bcrypt (cost factor 12+)

Include Anti-Patterns

Tell the AI what NOT to do:

## Avoid These Patterns

- Never use `any` type; use `unknown` with type guards
- Don't use `var`; prefer `const`, then `let`
- Avoid default exports; use named exports
- Don't mutate state directly; use immutable patterns

Provide Context

Explain WHY, not just WHAT:

## Why These Choices

We use Zustand instead of Redux because:

- Smaller bundle size for our performance requirements
- Simpler API for our team's experience level
- Better TypeScript inference

This context helps you make consistent recommendations.

Leveraging Prompt Inheritance

Create Reusable Base Prompts

# Base: typescript-strict

System Prompt:

- Use TypeScript strict mode conventions
- Explicit return types for all exported functions
- No implicit any
- Prefer interfaces over type aliases for objects

Override Only What Changes

Child prompts should add, not repeat:

# Child: react-typescript (extends typescript-strict)

System Prompt:

# React-Specific Additions

- Use functional components with hooks
- Prefer composition over inheritance
- (TypeScript rules are inherited automatically)

Document Inheritance

Add comments explaining the chain:

# Prompt: react-testing

# Extends: react-typescript -> typescript-strict

# Purpose: React component testing with Vitest

Managing Prompt Preferences

Match Team Conventions

Align preferences with your team's style guide:

{
  "verbosity": "balanced",
  "codeStyle": "documented",
  "responseFormat": "full"
}

Project-Specific Preferences

Different projects may need different settings:

Project	Verbosity	Code Style
Greenfield	Detailed	Documented
Maintenance	Concise	Minimal
Documentation	Detailed	Verbose

Auto-Activation Patterns

Project Configuration

Create .nexus/config.json in your project root:

{
  "defaultPrompt": "my-project-prompt",
  "autoActivate": true
}

Monorepo Setup

For monorepos, configure per-package:

my-monorepo/
├── .nexus/config.json          # Base config
├── packages/
│   ├── frontend/
│   │   └── .nexus/config.json  # Frontend-specific prompt
│   └── backend/
│       └── .nexus/config.json  # Backend-specific prompt

Memory Integration

Prompt-Specific Learning

When saving memories, associate them with prompts:

// Memory is automatically tagged with active prompts
save -
  memory({
    title: "React Query caching pattern",
    content: "...",
    // promptId is automatically added based on active prompts
  });

Recall by Prompt

Query memories in prompt context:

recall -
  memories({
    query: "caching patterns",
    // Automatically prioritizes memories from active prompts
  });

Performance Tips

Limit Active Prompts

Each active prompt adds to context size. Aim for 2-3 active prompts maximum.

Use Focused Libraries

Include only necessary documentation:

Good: react, react-query, zod
Bad: react, vue, angular, svelte, solid... (everything)

Streamline System Prompts

Keep prompts under 2000 tokens for optimal performance.

Troubleshooting Prompts

Unexpected Behavior

1. Check active prompts: nexus prompt active
2. Review merged configuration
3. Verify priority order
4. Check for conflicting instructions

Prompt Not Taking Effect

1. Ensure prompt is activated
2. Check inheritance chain
3. Verify you're authenticated
4. Try deactivating/reactivating

Context Overload

If responses seem unfocused:

1. Reduce number of active prompts
2. Trim system prompts
3. Remove unnecessary libraries
4. Use more specific prompts

Common Patterns

The "Project Onboarding" Prompt

# System Prompt

When starting work on this project:

1. Read README.md for project overview
2. Check package.json for dependencies
3. Review .nexus/config.json for project conventions
4. Look for existing patterns in src/ before creating new ones

Key conventions:

- [Project-specific rules here]

The "Code Review" Prompt

# System Prompt

When reviewing code:

1. Check for type safety issues
2. Verify error handling
3. Look for performance concerns
4. Ensure accessibility compliance
5. Validate test coverage

Output format:

- List issues by severity (Critical > High > Medium > Low)
- Include line numbers
- Provide fix suggestions

The "Documentation" Prompt

# System Prompt

When writing documentation:

1. Start with a one-line summary
2. Include usage examples
3. Document all parameters
4. Add error scenarios
5. Include related functions/components

Style:

- Use present tense
- Be concise but complete
- Include TypeScript types in examples

Next Steps

• Prompt CLI Commands - Full CLI reference
• MCP Tools - Programmatic prompt management
• Creating Prompts - Build custom prompts