Agent Engineering Starter Kit

## Overview
Agent Engineering Starter Kit gives builders three focused skills to ship more capable, efficient AI agents: context-optimization to reduce token waste, tool-design to build clean agent APIs, and prompt-engineering to write prompts that actually work.

## Quick Start
1. Install the plugin in your OpenClaw workspace: `npx skillstore add @agent-engineering-starter-kit`
2. Use **context-optimization** to audit your agent's memory layout — reduce token costs by partitioning context, applying KV-cache optimization, and setting context budgets before your first deployment.
3. Use **tool-design** to consolidate and simplify tool interfaces your agent exposes — apply architectural reduction principles so agents call tools safely and consistently.

## Key Commands
- `optimize context` — triggers context-optimization: improves context efficiency, masks irrelevant observations, extends effective context capacity
- `design tool` / `consolidate tools` — triggers tool-design: architects clean tool APIs with fewer sharp edges
- `prompt engineering` / `better prompts` — triggers prompt-engineering: applies chain-of-thought, few-shot, and system prompt techniques for Claude, GPT-4, Gemini and other models

## Tips
- Start with **prompt-engineering** to nail your system prompt, then apply **context-optimization** to trim what the agent actually needs to remember — together they cut token costs while keeping accuracy.
- Pair **tool-design** with context-optimization when your agent uses many tools: consolidating tool surface area is one of the fastest ways to reduce context pressure.
- Use all three skills in sequence when onboarding a new agent project: prompt structure → context budget → tool API design.