Skip to content

Latest commit

 

History

History
111 lines (90 loc) · 7.05 KB

README.md

File metadata and controls

111 lines (90 loc) · 7.05 KB

Documentation

This repository is the Codebase Agentic Readiness Framework. It says what makes a codebase safe for AI coding agents, and it holds tools that check a codebase. This page is the one router for the documentation. Each page under docs/ is linked below, once. The engine keeps its own docs tree beside its code; its router is under Components.

To audit a repository, open Claude Code at the root of this repository and say "Audit <repo-path> and write the output to <output-path>". To fix the findings, say "Start implementing the fixes on <repo-path> based on <audit-json-path>".

Tutorials

Page What it teaches
First audit on a fixture One guided audit on a small sample repository

How-to guides

Page What it does
Contribute a change Move one change from a work branch into dev through a pull request
Set up the agent tools Give your coding agent the code index, or skip it; the one-time offer agents make
Install the skills Put both skills from a release into a project, with no clone
Run an audit Produce a confirmed report through the agent and the audit recipe
Implement an audit report Apply the confirmed findings, one theme per branch
Resume an interrupted run Continue an audit or an implementation after a stop
Add a stack profile Add support for a new toolchain
Run the evals Evaluate the recipes headlessly

Reference

Page What it lists
Audit master list Every audit check by area: what passes, why it matters, which playbook fixes it
Implement master list Every playbook by theme: its goal, what it changes per check, what it never does
Tools CLI Each tool's synopsis, inputs, outputs and exit codes
Versions What each version number means, where it is written, and which reports and states the tools read
Harness notes What differs between Claude Code, opencode, Codex and Cursor
Checks to framework map Each check, its requirement and its framework section
Evaluation results Fixture-sweep and real-repository results
Self-audit This repository's own Phase 1 score and the CI ratchet
Development tooling Each tool that checks this repository, its command, its CI job and its pin
Branch protection The branch model, and the rulesets on dev and main
Licensing Which licence covers which files
Glossary The terms of the recipes and tools

Explanation

Page What it explains
Explanation index The two tracks below, in reading order
Purpose and non-goals What the recipes are, and what they refuse to be
How the recipes work The stages and decisions, for evaluators
Trust and safety What the recipes may do to a repository and a machine
Tool contract The contract that each tool follows
Report schema The schema, its versions and the engine mapping
Check catalogue and rubrics The catalogue, the methods and the rubric anatomy
Stack profiles How profiles adapt checks and playbooks
Evaluation The corpus, the metrics and the models
Architecture A map of the repository's parts and what each reads and writes

The framework

The framework is normative. If a tool disagrees with it, the tool is wrong.

Page What it holds
Overview What an agentic-ready codebase is
Phased approach The four phases; the recipes implement Phase 1
Brownfield track How to make a legacy codebase agentic-ready
Tool ecosystem Index tools and other tools the framework uses
Risks and mitigations The risk register
References Sources
Framework glossary The framework's own terms
spec/ The numbered requirements (REQ-...) for a repository
greenfield/ The recipe and templates for a new project

Decisions

The ADR index lists every decision of record, in all five groups. ADRs live beside the component they govern (ADR 0001). The ADR policy says how to add, change and delete one.

Components

Component Entry point
The audit recipe, its checks, schemas and tools audit/README.md
The implement recipe, its playbooks and tools implement/README.md
The deterministic audit engine (own docs tree) audit/engine/docs/architecture/README.md

The worked example shows a real run: two audits of one repository, before and after the implement recipe.

Writing docs

Page Use
Communication rules Plain-language rules, diagrams, and how to report and ask for decisions
Style guide The authoring rules for every page
Tutorial template Start a tutorial
How-to template Start a how-to guide
Explanation template Start an explanation page
Reference template Start a reference page
Architecture doc template Start an architecture page
ADR policy How to add, change and delete a decision record
ADR template Start a decision record

Gates and commit rules: AGENTS.md. Setup and contributing: CONTRIBUTING.md. Changes: CHANGELOG.md. Releases: RELEASE_PROCESS.md.