AI Assisted Coding Governance — AI skill for Claude Code

AI-Assisted Coding Governance

This repository provides policy and guidance for AI-assisted software development for yourself or clients. This repository is documentation only — there is no application code, no build, and no test runner.

---

Two tracks, same rules

The same governance is maintained for two different readers - AI agents and people:

• `ai-docs/` — for AI agents (coding and content work). Terse, imperative, rule-shaped. These are the files you copy into a client project.
• `human-docs/` — for people. Onboarding, the full developer guideline, and the shape a client's own AI policy takes.

Alongside them, **[`testing/`](./testing/)** holds the plan for verifying that this package actually works — that it installs correctly, and that its rules change how an agent behaves. It is a third directory rather than a third track: it ships to nobody and is not onboarding material. See [`testing/Governance-Test-Plan.md`](./testing/Governance-Test-Plan.md) for the plan and [`testing/coverage-matrix.md`](./testing/coverage-matrix.md) for what each scenario currently scores; the dated record behind both — every run, and every revision to the plan — is in [`testing/run-log.md`](./testing/run-log.md) and [`testing/test-plan-changes.md`](./testing/test-plan-changes.md).

**[`docs/`](./docs/)** is a fourth, on the same footing: the repo's own documentation, one subfolder per kind of document. Today that is [`docs/proposals/`](./docs/proposals/) — design proposals for changes to the package that **have not been made yet**. Nothing there describes how the package works today — read it to see what is under discussion and why, and read the owning file for current state. [`CHANGELOG.md`](./CHANGELOG.md) records what actually shipped.

Adopting this in a client project

**The same package installs the same way for everyone.** Every install lands two entry files — `AGENTS.md` and `CLAUDE.md` — plus an `ai-governance/` rules direc