Welcome

The open engineering harness for coding agents — train repeatable behavior, not model weights.

The Agents Playbook is an open engineering harness for coding agents. It turns rules, prompts, memory, evals, gates, and review patterns into repeatable repository behavior for agents like Claude, Cursor, Copilot, and Codex — without changing or fine-tuning the underlying model.

How it's organized

A matrix of 6 pillars × 6 SDLC phases. See the matrix for the cross-reference.

pillars/
  architecture/        # ADR, RFC, modular monorepo, contracts, errors, ...
  security/            # RBAC, vault, audit, compliance, AI/LLM safety, ...
  ui-ux/               # tokens, primitives, intl, a11y, whitelabel, ...
  quality/             # tests, gates, sanity, observability, FinOps, ...
  governance/          # PR intent, merge rules, tombstones, ...
  ai-collaboration/    # CLAUDE.md, MEMORY, sub-agents, slash commands, ...

phases/
  01-discover/   02-design/   03-build/
  04-test/       05-ship/     06-operate/

templates/    # ADR, RFC, PR-intent, CLAUDE.md, AGENTS.md, MEMORY.md
prompts/      # system, sub-agent, slash-command bodies
scripts/      # gate reference implementations (Node, zero deps)

Dual mode

Every doc has two layers:

TL;DR (human) — one paragraph for the linear reader.
For agents — structured sections, fixed shape, optimised for RAG retrieval and system-prompt injection.

Start here

Goal	Read
Adopt in a new project	Getting started
Review an agent-authored pull request	Turn review rules into executable evidence
Set non-negotiables for agents	CLAUDE.md template
Design a package boundary	Architecture · Universal
Add an ADR or RFC	ADR template · RFC template
Wire quality gates	Quality gates
Train an agent on lessons	MEMORY template
Multi-agent merge	Governance

The eight non-negotiables

The irreducible kernel is eight rules (not ten). Landing samples and templates use the same eight. If an agent breaks one of these, fail the PR.

Typed boundaries. Every external input is parsed by a runtime schema. No any.
Named exports only. Predictable refactors, predictable agent edits.
Typed error hierarchy with stable codes. AppError subclasses with <NS>_<CODE> constants.
Centralized logger. createLogger(tag); never console.log in shipped code.
ADR before architecture change. RFC before breaking a public contract.
Ship complete or don't ship. No TODO/FIXME/not implemented in shipped surfaces.
Merges sum work, never subtract. PR intent manifest; removes: justified.
Tokens, intl, primitives. No raw values in user-facing surfaces.

Each is fully spec'd across the pillars and enforced by gate scripts.

In the AgentsKit ecosystem

Playbook is the discipline product. Pair it with:

AgentsKit — foundation library and runtime
AgentsKit Chat — product chat experience (dogfood notes)
Doc Bridge — deterministic documentation handoffs
Registry — copy-ready agent starting points
AKOS — enterprise operation

For agents reading this

Start with the concise For agents route to choose the smallest useful surface.
The full bundle in one file: /llms-full.txt.
Site map: /llms.txt.
Raw markdown for any doc: replace /docs/<path> with /raw/<path>.md.
ZIP of all docs: /playbook-bundle.zip.