Harness engineering is the discipline of building agent-ready environments — the systematic practice of configuring, structuring, and optimizing a project so that AI agents can operate effectively within it.

#Definition

Harness engineering encompasses:

#Relationship to Other Disciplines

Harness engineering builds on and complements:

#The Agent-Ready State

A project achieves agent-ready status when an agent can:

  1. Enter the repository and understand its purpose, structure, and conventions through CLAUDE.md or equivalent
  2. Execute tasks without requiring constant human guidance
  3. Follow established patterns for commits, testing, and documentation
  4. Access external tools and data through standardized interfaces
  5. Record failures and contribute to the project's evolution