Prompting for C# Development

Overview

Prompt quality directly affects generated code quality. Strong prompts include context, constraints, and acceptance criteria.

Prompt Structure

Use a consistent format:

sequenceDiagram
    participant Dev as Developer
    participant AI as AI Assistant
    participant CI as CI Pipeline

    Dev->>AI: Provide task + context + constraints
    AI-->>Dev: Draft code and tests
    Dev->>CI: Run build and tests
    CI-->>Dev: Pass or fail feedback
    Dev->>AI: Request targeted fixes if needed

Practical Example - Good Prompt for API Handler

Task: Add idempotency support to POST /payments.
Context: ASP.NET Core 9, MediatR, SQL Server, existing table Payments(Id, RequestId, Amount, Status).
Constraints: Keep response schema unchanged. Reject duplicate RequestId with 200 and existing result.
Acceptance: Unit tests for new and duplicate requests. No N+1 queries. Structured logs include request_id.

Expected output quality improves because the model has explicit boundaries.

C# Example - Contract and Handler Skeleton

public sealed record CreatePaymentCommand(string RequestId, decimal Amount) : IRequest<CreatePaymentResult>;

public sealed record CreatePaymentResult(Guid PaymentId, string Status);

public sealed class CreatePaymentHandler : IRequestHandler<CreatePaymentCommand, CreatePaymentResult>
{
    public async Task<CreatePaymentResult> Handle(CreatePaymentCommand request, CancellationToken ct)
    {
        // Idempotency behavior belongs here: lookup by RequestId before insert.
        throw new NotImplementedException();
    }
}

Production Example - Prompt Templates in Repositories

Many teams keep prompt templates in version control under a folder such as docs/ai-prompts. Templates are reviewed like code and updated when architecture patterns change.

Common Pitfalls

Key Takeaways