Documentation
Overview
harness-miniis a minimal outer harness for developers' long coding tasks: repo conventions + skills + sub-agents + distilled docs, with only thin shell glue as code. It keeps work on-task, context-light, and resumable across sessions.
The contract is files + shell. Any coding agent that can read the repo and run commands can drive it — no vendor-specific runtime. The repository is the system of record: anything not in a version-controlled file does not exist to the agent.
$ git clone https://github.com/harness-mini/harness-mini.git$ bash harness-mini/init.sh /path/to/your/projectThe 40% line
The smart/dumb contract — the intellectual core. How every agent keeps its context in the smart zone.
Principles
The Mini delete-first constraint, the golden principles, and the clean-code / refactor quality spine.
Architecture
Repository layout, the lifecycle FSM, the horizontal+vertical coding contract, and new-vs-existing install.
References distilled in the harness
- Anthropic — Effective harnesses for long-running agents
- Anthropic — Harness design for long-running apps
- Anthropic — Scaling managed agents (decoupling brain/hands/session; stale assumptions)
- OpenAI — Harness engineering (leveraging Codex)
- Anthropic — The Founder's Playbook