These pages document master, which is unreleased and in development. The Quick Start installs the latest stable release; anything newer than that tag is marked in the text.
Evidence & Influences
External reviews, scorecards, and the sources that shaped this harness. Moved out of the README front fold (#211, D11) so that above the fold carries only objective signals — release, CI, and install paths.
External scorecards
Section titled “External scorecards”Independently scored 5.0 / 5.0 on OpenAI Harness Engineering and 10.0 / 10.0 on Claude Code Best Practice (February 2026) — see the scorecards.
These are point-in-time external assessments of the harness design, not continuous measurements; the continuously enforced signals are the CI gates (test suites, sync check, doc-accuracy drift gate).
Influences
Section titled “Influences”- Stop Fighting AI Agents and Build a Reusable Multi-Agent Dev Environment — the project’s origin story: 13+ real build sessions and the six recurring failure patterns every rule here traces back to.
- Spec-Driven Development vs Code Copilot Team — how this harness relates to GitHub’s Spec Kit: SDD defines what to build; this project defines how to behave while building it.