AndysTMC

Knowledge Architecture — Development skill for Claude Code

Development community

A typed system for project knowledge that humans and coding agents share.

How to install Knowledge Architecture

This entry records only its repository, not the path inside it, so there is no exact command to give. Open AndysTMC/knowledge-architecture and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Knowledge Architecture does

A typed system for project knowledge that humans and coding agents share.

Alternatives in Development

  • Code Review Graph — Local knowledge graph for Claude Code 3.7k ★
  • Reverse API Engineer — Reverse engineer any website into a typed API client 1.9k ★
  • Cq — An open standard for shared agent learning 1.3k ★

README

Knowledge Architecture

**0.2.0** — a typed system for project knowledge that humans and coding agents share. Latest published release: **v0.2.0**. Published tags are immutable.

Most repositories that coding agents work in end up with conflicting `AGENTS.md`, `CLAUDE.md`, TODO, and notes files, and nothing tells an agent which of them is true. This project gives each kind of knowledge one home: how to operate, what we decided, what we are doing now. Each kind has an authority level, and a small linter keeps the result honest.

  • Kernel: the one-page version. Start here.
  • Full specification: the model, the layout, and the edge cases.
  • Implementation prompt: hand it to an agent to apply the spec to a repository.
  • scripts/lint_knowledge.py: a single file that checks the tree, scaffolds a new repository, and gates decision changes in pull requests.

This repository [applies the architecture to itself](docs/this-repo.md).

Requirements

Python **3.10+**, standard library only. Git is needed for the decision gate. There is nothing to install.

Use it in a repository

Pick the case that matches the target repository. Always use the same release (`v0.2.0` below) for the linter and the documents.

1. New repository, or no agent files yet

From the target repository, run:

curl -fsSL --create-dirs -o scripts/lint_knowledge.py \
  https://raw.githubusercontent.com/AndysTMC/knowledge-architecture/v0.2.0/scripts/lint_knowledge.py
python3 scripts/lint_knowledge.py --init --test "pytest"
python3 scripts/lint_knowledge.py --strict

Replace `pytest` with the real test command; `--install`, `--dev`, and `--lint-command` are optional. Init creates a README if there is none, an `AGENTS.md` hub, the three tool pointers (`CLAUDE.md`, `GEMINI.md`, `.github/copilot-instructions.md`), and the vendored linter. It never creates empty `docs/` folders, and it never overw