/ai-prompting

name: ai-prompting
description: Effective communication strategies for AI-assisted development. Learn context-first prompting, phased interactions, iterative refinement, and validation techniques to get better results from Claude and other AI coding assistants.
license: MIT
compatibility: Claude Code, Cursor, VS Code, Windsurf
metadata:
  category: methodology
  complexity: beginner
  author: Kiro Team
  version: "1.0.0"

AI Prompting Strategies

Master the art of communicating with AI coding assistants to get better results faster. These strategies are optimized for spec-driven development but apply broadly to AI collaboration.

When to Use This Skill

Use these prompting strategies when:

• Working with Claude Code, Cursor, or other AI assistants
• Creating specs through AI collaboration
• Getting inconsistent or low-quality AI responses
• Need to improve AI output accuracy
• Want faster iteration cycles

Core Strategies

Strategy 1: Context-First Prompting

Always provide sufficient context before making requests.

**Poor Approach:**

Create requirements for a user profile feature.

**Better Approach:**

I'm working on a web application for a fitness tracking platform. We need to add user profile functionality where users can manage their personal information and fitness goals.

Context:
- Technology: React frontend, Node.js backend
- User base: Health-conscious individuals, age 18-65
- Key constraint: Must comply with GDPR for EU users
- Integration: Will connect with existing authentication system

Please help me create requirements for the user profile feature.

**Why It Works:**

• Provides domain context for better decisions
• Identifies technical constraints early
• Clarifies compliance requirements
• Enables more relevant suggestions

Strategy 2: Phased Interaction

Work through spec phases sequentially. Complete each phase before moving to the next.

**Phase 1 - Requirements:**

Let's start with the requirements phase for [feature name].

Current situation: [describe current state]
Problem to solve: [describe the problem]
Users affected: [describe user types]
Success criteria: [how we'll know it works]

Please help me develop comprehensive requirements using the EARS format.

**Phase 2 - Design (after requirements approved):**

Now that we have clear requirements, let's create the technical design.

Requirements summary: [key requirements]
Technical context: [architecture, frameworks, patterns]
Constraints: [performance, scalability, security]

Please propose a technical design that addresses these requirements.

**Phase 3 - Tasks (after design approved):**

With the design finalized, let's break this into implementation tasks.

Design summary: [key components and interactions]
Team context: [team size, skill levels]
Dependencies: [what must be built first]

Please create a sequenced task breakdown for implementation.

Strategy 3: Iterative Refinement

Treat spec development as conversation, not single requests.

**Initial Request:**

Help me define requirements for email notification preferences.

**Refinement Round 1:**

Great start! Let's refine a few areas:
1. For notification frequency, can we add daily digest option?
2. How should we handle changing preferences during pending notifications?
3. Can you elaborate on the unsubscribe requirement for GDPR compliance?

**Refinement Round 2:**

Perfect. Now let's add requirements for:
- Mobile push notifications (in addition to email)
- Notification history (last 30 days)
- Per-notification-type controls (not just global on/off)

Strategy 4: Example-Driven Prompting

Provide concrete examples of what you want.

**For Requirements:**

I need acceptance criteria for a file upload feature. Use the EARS format like this example:

Good example from our auth feature:
"WHEN a user enters valid credentials THEN the system SHALL authenticate within 2 seconds"

Avoid vague requirements like:
"System should handle file uploads efficiently"

Focus on specific, testable criteria for:
- File size limits
- Supported file types
- Upload progress indication
- Error handling

**For Design:**

Create a component architecture. Follow this existing pattern:

[Reference existing architecture]

Key elements to include:
- Component responsibilities
- Data flow
- API boundaries
- Error handling paths

Strategy 5: Constraint-Explicit Prompting

Make all constraints explicit. Don't assume AI knows your limitations.

Design a caching strategy for product catalog data.

Explicit constraints:
- Infrastructure: AWS with Redis, PostgreSQL
- Performance: API response < 200ms for cached data
- Scale: 10,000 products, 1,000 concurrent users
- Budget: Cache cost < $100/month
- Freshness: Updates visible within 5 minutes
- Maintenance: 2-person ops team

Flexibility allowed:
- Cache invalidation strategy (time or event-based)
- Cache key structure (optimize as needed)
- Failover approach (as long as reliable)

Strategy 6: Role-Based Prompting

Frame requests from specific perspectives.

**Product Owner Perspective:**

As a product owner defining checkout requirements:
- Business goals: Reduce cart abandonment
- User value: Smooth, trustworthy purchase experience
- Success metrics: Checkout completion rate > 80%

What requirements should I capture?

**Technical Lead Perspective:**

As tech lead designing a notification system:
- Integrates with existing microservices
- Handles 100k notifications/day with room to grow
- Maintains health if notification service fails
- Aligns with event-driven architecture

What design approach would you recommend?

**Developer Perspective:**

As a mid-level developer implementing this:
- Need clear tasks (2-4 hours each)
- Explicit dependencies between tasks
- Guidance on testing approach
- References to existing code patterns

Can you break down the implementation accordingly?

Strategy 7: Val