hello-world

Hello World — Harness Engineering (Basic)

A minimal project showing what harness engineering looks like at the basic adoption level.

What is this?

A tiny greeting library managed by harness. It demonstrates the foundation: configuration, validation, and agent context — the building blocks every harness project uses.

Try it

cd examples/hello-world
npm install
harness validate

You should see validation pass. This confirms:

• harness.config.json is valid
• AGENTS.md exists and is readable
• Layer definitions are syntactically correct

What just happened?

harness validate checked your project's configuration and structure. At the basic level, this means:

• The config file parses correctly and has required fields
• The AGENTS.md knowledge map exists where the config says it should
• Layer definitions (if any) are well-formed

No linting rules are enforced at the basic level — that comes in the task-api example.

Explore the config

Open harness.config.json:

{
  "version": 1,
  "name": "harness-hello-world",
  "layers": [...],
  "agentsMapPath": "./AGENTS.md",
  "docsDir": "./docs",
  "template": { "level": "basic", "version": 1 }
}

• layers — Defines the architectural layers. Even at basic level, you declare them. They're enforced starting at intermediate level.
• agentsMapPath — Where AI agents look for project context.
• template.level — Which harness adoption level this project uses.

What does sample state look like?

Check .harness.example/ to see what a project's state directory looks like after a few sessions:

• state.json — Current position, progress, decisions
• learnings.md — Institutional knowledge captured over time

In a real project, this would be .harness/ (not .harness.example/).

Next

Ready for layer enforcement, ESLint rules, and personas? Try the task-api example.