Introduction

Harness Kit helps teams install repeatable AI engineering guardrails with small, reviewable diffs. Instead of manually copying prompts, rules, and agent configs between repositories, you define a baseline once and keep it consistent through CLI commands.

Why teams adopt it

Reduce setup drift across projects and contributors.
Install capabilities incrementally, not all at once.
Keep generated artifacts explicit and version controlled.
Scale team conventions without copy-paste setup debt.

What changes in your repository

Baseline guidance files such as CLAUDE.md and AGENTS.md.
A tracked harness state file (harness.json) listing installed bundles.
Optional automation artifacts, for example rules, skills, hooks, and MCP settings.

Core concepts

Bundle: one installable capability (for example tdd, nextjs, security-review) that contributes project artifacts.
Optional automation assets such as hooks, memories, and role-focused instructions.
Bundle docs category: docs grouping for discovery (workflow, stack, techstack).
CLI category filter: install-role filter for harness-kit list --category (for example workflow-preset, memory, browser).

Fastest onboarding path

Step 1: Open Quickstart and run the copy-paste flow.
Step 2: Use init to create the baseline.
Step 3: Add one bundle, verify with status, then commit.

Initialize

Validate generated harness state

Add one workflow bundle

Success criteria

Your repo contains expected harness files and no unknown setup drift.
Teammates can run the same commands and get equivalent output.
Each added capability lands as a small, auditable diff.

Official GitHub repository

Source code and project history are publicly available at github.com/timezlab/harness-kit.