AI-assisted commands for architects practising the Continuous Architecture Framework (CAF).
This repository provides 25 ready-to-use commands for AI coding assistants — Claude Code, GitHub Copilot, Gemini — that operationalise the full CAF across its six views, five practices, and five rituals. Each command produces structured, traceable architecture artifacts aligned with the CAF manifesto.
This repository is a companion to the Continuous Architecture Toolkit. The toolkit defines the framework. This repo puts it into practice with AI.
Architecture frameworks often live in documentation. Practitioners read them, then face a blank page when starting a real engagement.
These commands close that gap. They embed CAF thinking directly into your AI assistant: the right questions, the right outputs, the right quality gates — for each phase of the product lifecycle, across every CAF view.
The approach is inspired by ArcKit, adapted and extended to cover the full CAF, including the Organisation and Enterprise Decomposition views that most AI toolkits overlook.
# 1. Clone this repository
git clone https://github.com/continuous-architecture/ai-commands.git
# 2. Copy the commands into your project (Claude Code)
cp -r ai-commands/.claude/commands/ my-project/.claude/commands/
# 3. Start Claude Code
cd my-project
claude
# 4. Initialise your CAF project
/caf.init Payment modernisation programme, retail bank, 80 engineers| Command | What it produces |
|---|---|
caf.init |
Project structure, CLAUDE.md, seeded principles file |
caf.review |
Cross-view RAG health assessment for Architecture Review Board |
caf.report |
2-page executive report for CTO / Programme Director |
caf.self-assessment |
Maturity scoring across all views, practices, and rituals + improvement roadmap |
| Command | What it produces |
|---|---|
caf.xo-jtbd |
Jobs-to-be-done analysis — performer, jobs, stages, needs, circumstances |
caf.xo-personas |
Persona cards + experience map + architecture sensitivity matrix |
| Command | What it produces |
|---|---|
caf.product-runway |
Architecture runway — tech radar, enabler backlog, consolidated roadmap |
caf.product-adr |
Individual Architecture Decision Record (full CAF 5-step process) |
caf.product-scoping |
Scoping 360 — problem framing, team composition, architecture approach, delivery plan |
| Command | What it produces |
|---|---|
caf.tech-principles |
Assessment of the 4 CAF technology principles with RAG + DORA metrics |
caf.tech-fitness |
Fitness functions register + board + implementation roadmap |
caf.tech-eventstorming |
Event storming facilitation package — brief, domain events, bounded context candidates, aggregates |
| Command | What it produces |
|---|---|
caf.ops-capabilities |
VCAP capability assessment — values, capabilities, assets, processes |
caf.ops-readiness |
Operational readiness gate — observability, failure, security, capacity, deployment |
caf.ops-digital |
Digital operating model — automation, AI/ML, self-care, developer experience |
| Command | What it produces |
|---|---|
caf.org-team-design |
Team topology (Team Topologies typology), cognitive load assessment, dependency map |
caf.org-interaction-modes |
Team API cards, interaction mode register (Collaboration / X-as-a-Service / Facilitating) |
caf.org-inverse-conway |
Conway's Law audit, coupling heatmap, action plan — RAG status for ARB |
| Command | What it produces |
|---|---|
caf.ed-operating-unit-map |
OU hierarchy, decomposition criteria, cross-OU dependencies |
caf.ed-product-portfolio |
Product inventory per OU, platform candidates, lifecycle heat map |
caf.ed-domain-map |
Bounded context map (DDD), OU–domain alignment, core domain identification |
| Command | What it produces |
|---|---|
caf.practice-adr |
ADR index + decision health check (stale, orphaned, coverage gaps) |
| Command | What it produces |
|---|---|
caf.ritual-kata |
Architecture kata facilitation package — brief, format, session output |
caf.ritual-peer-review |
Peer review request + feedback register + dissent log |
caf.ritual-weekly |
Weekly architecture meeting agenda + minutes template |
You need one of the following AI assistants:
- Claude Code — recommended, full agent support
- GitHub Copilot with custom instructions support
- Gemini CLI
You do not need to know how to code. These are text files.
git clone https://github.com/continuous-architecture/ai-commands.gitClaude Code:
cp -r ai-commands/.claude/commands/ my-project/.claude/commands/GitHub Copilot:
cp -r ai-commands/.copilot/ my-project/.copilot/Gemini:
cp -r ai-commands/.gemini/ my-project/.gemini/cd my-project
claude
/caf.init [your programme context in plain language]Each command accepts a plain-language description of your context as input.
/caf.xo-jtbd
Corporate treasury management platform, targeting mid-size companies, 3 personas
/caf.org-team-design
E-commerce replatforming, migrating from monolith, 6 teams, 80 engineers
/caf.self-assessment
We are 6 months into our CAF adoption, team of 12 architects across 4 programmes
The command reads any existing project artifacts, asks for clarification if critical context is missing, and produces a structured Markdown document.
caf.init
│
├── caf.xo-jtbd ──→ caf.xo-personas
│
├── caf.ed-operating-unit-map ──→ caf.ed-product-portfolio ──→ caf.ed-domain-map
│
└── caf.org-team-design ──→ caf.org-interaction-modes ──→ caf.org-inverse-conway
caf.product-runway ──→ caf.product-adr (per decision)
caf.tech-principles ──→ caf.tech-fitness
caf.ritual-weekly (every week)
caf.ops-readiness ──→ caf.review ──→ caf.report
caf.self-assessment ──→ caf.review ──→ caf.report
caf.tech-eventstorming [domain] ← when exploring a new domain
caf.ritual-kata [challenge] ← quarterly or on a hard problem
caf.ritual-peer-review [ADR] ← before committing a significant decision
caf.product-scoping ← at the start of a new product or team
my-project/
├── CLAUDE.md ← auto-context for Claude Code (from caf.init)
├── project/
│ ├── principles.md
│ ├── requirements.md
│ ├── architecture-runway.md
│ ├── caf-review-YYYY-MM-DD.md
│ ├── executive-report-YYYY-MM-DD.md
│ └── self-assessment-YYYY-MM-DD.md
│
├── experience-objectives/
│ ├── jtbd.md
│ └── personas.md
│
├── product/
│ ├── portfolio.md
│ └── scoping-360.md
│
├── technology/
│ ├── tech-principles.md
│ ├── fitness-functions.md
│ ├── event-storming-[domain].md
│ └── adrs/
│ ├── README.md ← ADR index (from caf.practice-adr)
│ ├── ADR-001-[title].md
│ └── ...
│
├── operations/
│ ├── capabilities.md
│ ├── readiness.md
│ └── digital-operating-model.md
│
├── organisation/
│ ├── team-design.md
│ ├── interaction-modes.md
│ └── inverse-conway-audit.md
│
├── enterprise/
│ ├── operating-unit-map.md
│ ├── product-portfolio-map.md
│ └── domain-map.md
│
└── rituals/
├── weekly-arch-YYYY-MM-DD.md
├── kata-[topic]-YYYY-MM-DD.md
└── peer-review-[topic]-YYYY-MM-DD.md
Alignment + Autonomy > Control. Commands produce artifacts that help teams align without imposing top-down control. Every output is a starting point for conversation, not a constraint.
Grounded in real problems. Each command reads existing project context before generating anything. It does not produce generic content — it reasons from your specific situation.
AI-agnostic by design. The same commands work across Claude Code, Copilot, and Gemini. Output format is identical regardless of the underlying model. Your artifacts are portable.
Traceable. Every command produces quality gates that verify traceability back to principles, requirements, and other artifacts. Architecture decisions do not appear from nowhere.
A command is a Markdown file. If you have a CAF practice that could be assisted by AI, you can contribute it — no coding experience required.
See CONTRIBUTING.md for the command template and review process.
Join the conversation on LinkedIn or at our events.
These commands draw on:
- Team Topologies — Skelton & Pais
- Domain-Driven Design — Evans / DDD-crew
- Open Agile Architecture — The Open Group
- Scaled Agile Framework (SAFe)
- Building Evolutionary Architectures — Ford, Parsons, Kua (fitness functions)
- ArcKit — inspiration for the command structure
Apache License 2.0 — see LICENSE.
Content is part of the Continuous Architecture open community.