api-and-interface-design

Files (1)

• SKILL.md

SKILL.md (first lines)

# API and Interface Design

## Overview

Design APIs and module boundaries with contract-first thinking. Apply Hyrum's Law (all observable behaviors will be depended on), the One-Version Rule (prefer one way to do something), and clear error semantics. Public interfaces deserve more care than internal ones.

## When to Use

- Designing REST or GraphQL API endpoints
- Creating module boundaries within an application
- Building SDKs, libraries, or shared packages
- Defining function signatures that cross module boundaries

## Core Principles

### Contract-First Design

Define the contract before implementation:
- Input shapes with validation rules
- Output shapes with all possible variants
- Error responses with clear semantics
- Versioning strategy

### Hyrum's Law

Every observable behavior of your API will eventually be depended on by someone. Design accordingly:
- Be explicit about what is guaranteed vs what is implementation detail
- Document only what you commit to maintaining
- Hide internal details behind stable interfaces

### One-Version Rule

Prefer one way to accomplish a given task:
- Avoid multiple endpoints that do the same thing with different syntax
- Avoid multiple function signatures for the same operation
- If you must provide convenience methods, make one canonical path clear

### Error Semantics

Errors should be:
- Specific enough to be actionable
- Consistent across the API surface
- Accompanied by enough context to debug
- Categorized: client errors (4xx) vs server errors (5xx)

### Boundary Validation

Validate all input at system boundaries:
- External API endpoints
- Module boundaries
- Database queries
- File parsing

## Process

### Step 1: Define the Contract

Write the API contract before any implementation:
- Endpoint paths and HTTP methods
- Request and response schemas
- Error response formats
- Pagination patterns if applicable

### Step 2: Validate the Contract

Review for:
- Consistency with existing API patterns
- Naming clarity and predictability
- Completeness of error cases
- Backward compatibility implications

### Step 3: Implement Against the Contract

- Implementation should match the contract exactly
- If the contract needs to change, update the contract first
- Add contract tests that verify implementation matches

### Step 4: Document the Public Surface

- Document only guaranteed behavior, not implementation details
- Include examples for common use cases
- Document error cases and recovery strategies

## Common Rationalizations

| Rationalization | Reality |
|---|---|
| "I can refine the API after it works" | Breaking changes after adoption are expensive and disruptive. |
| "This is an internal API, it does not matter" | Internal APIs become external APIs the moment another team depends on them. |
| "Adding more parameters is backward compatible" | Adding parameters increases the surface area that consumers may depend on. |

## Verification

- [ ] Contract is defined before implementation
- [ ] All error cases have defined responses
- [ ] Boundary validation covers all external inputs
- [ ] Public surface is documented with examples
- [ ] Contract tests verify implementation matches specification

## Anti-Rationalization Table

| Excuse | Counter |
|--------|---------|
| "I'll define the contract as I implement" | Without a contract, implementation drifts. Define it first to catch design flaws early. |
| "This is an internal API, Hyrum's Law doesn't apply" | Internal APIs become external the moment another team depends on them. Design accordingly. |
| "Adding one more endpoint is quick" | Every new endpoint increases the surface area consumers depend on. One-version rule prevents confusion. |
| "Error handling can be added later" | Inconsistent error semantics are hard to retrofit. Define error responses upfront. |
| "The contract tests are too much overhead" | Contract tests catch drift between spec and implementation. They prevent silent breaking changes. |

From v1truv1us/ai-eng-system, MIT.

More from v1truv1us/ai-eng-system