ECC Harness Methodology: Agent Performance Optimization

This skill distills Everything Claude Code's core methodology for building production-grade agentic systems: how to optimize token spend, orchestrate automated hooks, design resilient agent harnesses, and embed TDD into agent workflows.

Core Conceptual Model

The Agent Harness = { Token Budget } + { Hook Pipeline } + { Agent State Machine } + { Test Coverage }

An agent harness is a closed-loop system where:

• Tokens are your most constrained resource (cost, latency, context windows)
• Hooks are decision gates that intercept agent actions (pre/post execution, safety checks, state capture)
• Agents are task-specific policies that route work (specialized models, role-based instructions)
• Tests are your only proof the harness works end-to-end under real conditions

1. Token Optimization (Priority 1)

The Token Budget Framework