AIOS Workspace Template
PMatX3/mattis-aios-workspace-template
Sistema operacional de IA para rodar seu negócio com Claude como agente persistente
Ver no GitHubUm template de workspace estruturado que cria uma camada de contexto persistente para Claude Code, permitindo que o modelo funcione como parceiro de negócio informado. Inclui 12 comandos, 9 skills modulares, worktrees para sessões paralelas e templates prontos para automação com agentes.
Instalação
npx convez-ai PMatX3/mattis-aios-workspace-template1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Fundadores e consultores que precisam de IA com contexto de negócio
- Automação de workflows recorrentes (planejamento, execução, auditoria)
- Execução de tarefas em paralelo via worktrees sem conflitos de contexto
- Integração de agentes autônomos do Claude em GitHub Issues
Veredito

Template robusto que resolve o problema real de contexto persistente em sessões Claude. Bem estruturado, modular e production-ready para times que querem usar IA como ferramenta operacional genuína, não apenas chat. Recomendado para quem busca infraestrutura séria, não toy projects.
README
A tradução para o português deste README ainda está na fila.
AIOS Workspace Template
An AI Operating System for running your business with Claude Code.
AIOS is a structured workspace that gives Claude Code persistent context about you, your business, and your goals — so every session starts informed, not from zero. It's designed for founders, consultants, and operators who want an AI that functions as a genuine business partner, not a generic chat tool.
What This Is
Most AI interactions start with re-explaining everything. AIOS solves this by building a context layer Claude reads at the start of every session. Once set up:
- Claude knows your business, role, and current strategy
- You never re-explain who you are or what you're doing
- 12 structured commands handle common workflows (planning, executing, content, auditing)
- 9 skills extend Claude's capabilities (writing enforcement, agent architecture, scope checking)
- Worktrees let you run parallel Claude sessions (delivery, commercial, research) without conflicts
- GitHub issue templates enable autonomous Claude agents to work tickets without hand-holding
- Module installers add integrations (GHL, GitHub, Calendly, content pipeline) in minutes
Quick Start
# 1. Clone this template
git clone https://github.com/YOUR_USERNAME/mattis-aios-workspace-template my-workspace
cd my-workspace
# 2. Run setup (creates worktrees + installs shell aliases)
bash scripts/setup.sh
source ~/.zshrc
# 3. Fill in your context files
# Edit these 4 files with your business details:
# context/business-info.md
# context/personal-info.md
# context/strategy.md
# context/current-data.md
# OR: run the ContextOS module to have Claude interview you (see module-installs/context-os/)
# 4. Launch Claude
cs # launches Claude and runs /prime automatically
After step 4, Claude will read your context files and confirm it understands your business.
Folder Structure
.
├── CLAUDE.md # Master context file — always loaded, keep current
├── HISTORY.md # Workspace changelog — updated by /commit
├── DESIGN.md # Design system — fill in your brand colours, fonts, voice
├── .claude/
│ ├── commands/ # 12 slash commands
│ └── skills/ # 9 modular skill packs
├── context/ # Who you are, your business, strategy, metrics
│ ├── business-info.md
│ ├── personal-info.md
│ ├── strategy.md
│ ├── current-data.md
│ ├── decision-policy.md
│ └── import/ # Drop docs here for Claude to ingest
├── docs/ # Self-documenting workspace
│ ├── _index.md # Documentation routing index
│ └── _templates/ # System and integration doc templates
├── plans/ # Dated implementation plans
├── outputs/ # Deliverables and registers
├── artifacts/ # Project-specific outputs
├── reference/
│ ├── scorecard-metrics.md
│ ├── shell-aliases.md
│ └── templates/ # 6 delivery templates (SOW, UAT, runbook, etc.)
├── module-installs/ # 7 plug-and-play modules
├── scripts/
│ ├── setup.sh # One-time workspace setup
│ └── wt.sh # Worktree helper
└── .github/
└── ISSUE_TEMPLATE/
└── agent-task.md # Template for agent-ready GitHub issues
Commands
Run these inside a Claude Code session (after cs or claude):
| Command | What it does |
|---|---|
/prime | Loads your context and confirms Claude understands your business |
/create-plan [request] | Creates a detailed implementation plan in plans/ |
/implement [plan-path] | Executes a plan created by /create-plan |
/commit [message] | Saves work, updates docs and changelog, offers to push |
/install [module] | Installs an AIOS module from module-installs/ |
/brainstorm [topic] | Scans your workspace for automation opportunities |
/explore [idea] | Shapes an idea through 5-stage interactive Q&A |
/share [system] | Packages a system for sharing with installer |
/task-audit | Maps all tasks and scores automation potential |
/capture [idea] | Quick content idea capture into pipeline |
/develop [#ID] | Develops a content idea into a full concept |
/schedule | Interactive content scheduling session |
The ideation cycle: /brainstorm > /explore > /create-plan > /implement
Skills
Skills are modular capability packs that activate automatically when relevant:
| Skill | What it does |
|---|---|
generate-uat | UAT plans from GitHub repos (reads actual codebase) |
intent-check | AI delegation safety framework (4 modes) |
scope-check | Analyse client requests against your SOW/contract |
skill-creator | Create or update AgentSkills |
mcp-integration | Model Context Protocol server setup |
writing-style | Anti-AI-slop enforcement for all prose |
vibe-coding | Toolkit for non-technical builders (6 modes) |
build-or-buy | Agent deployment decisions using the 4:1 Ratio |
which-agent | Agent architecture classifier (3 modes) |
Shell Aliases
After running setup.sh, you get:
| Alias | Description |
|---|---|
cs | Launch Claude (safe mode — prompts for permission) and run /prime |
cr | Launch Claude (auto-approve) and run /prime |
wt | cd to workspace root |
wtd | cd to delivery worktree |
wtc | cd to commercial worktree |
wtr | cd to research worktree |
Parallel Worktrees
The workspace creates three isolated branches so you can run parallel Claude sessions without conflicts:
| Branch | Use for |
|---|---|
main | Planning, strategy, workspace maintenance |
delivery | Client work, UAT plans, runbooks, handover packs |
commercial | Proposals, content, outbound, pricing |
research | Analysis, strategy thinking, background research |
Open each branch in a separate terminal > run cs > Claude sessions work independently.
Customise It
The context files are where your business lives. Fill them in:
context/business-info.md— What your business does, who it serves, key products/servicescontext/personal-info.md— Your role, responsibilities, how the workspace helps youcontext/strategy.md— Current priorities, what success looks like, key decisionscontext/current-data.md— Key metrics, current state, data sourcesDESIGN.md— Your brand colours, fonts, animation style, CTA defaults
Then update CLAUDE.md to remove any generic text and reflect your actual business.
Tip: Use the ContextOS module (
/install module-installs/context-os) to have Claude interview you and build context files automatically.
Agent-Ready GitHub Issues
This workspace includes a convention for running Claude Code agents autonomously on GitHub issues.
How it works
- Write an issue using the Agent Task template (
.github/ISSUE_TEMPLATE/agent-task.md) - Add the
agent-readylabel - In a worktree terminal:
claude> paste the issue URL > agent reads and executes - Agent opens a PR > adds
human-reviewlabel - You review and merge
Labels
| Label | Meaning |
|---|---|
agent-ready | Fully specified; agent can pick it up |
agent-in-progress | Agent is working it |
needs-spec | Needs more detail before an agent can work it |
human-review | Work done; needs human sign-off |
Delivery Templates
Six production-tested templates in reference/templates/:
| Template | Use for |
|---|---|
audit-checklist.md | 7-day structured audit of an existing system |
sow-scope-acceptance.md | Statement of Work with acceptance criteria |
uat-plan.md | UAT plan with test cases and sign-off |
go-live-runbook.md | Go-live sequence, rollback plan, monitoring |
handover-pack.md | Ownership transfer documentation |
change-request.md | Scope/timeline/price change control |
Weekly Operating Rhythm
Every Monday (~90 min):
outputs/project-register.md— update project status, risks, next milestonereference/scorecard-metrics.md— fill this week's scorecard row- Check WIP count against your limit
- Publish one piece of content for your primary acquisition channel
- Send targeted outreach to ICP
- Unblock your team
Module Installers
module-installs/ contains plug-and-play modules that extend the workspace:
| Module | What it does | Setup time |
|---|---|---|
context-os/ | Guided interview to build all 4 context files | 30-45 min |
content-pipeline/ | Content capture, develop, schedule workflow with strategy workshop | 30-45 min |
github-os/ | GitHub commit and PR activity collection | 10-15 min |
ghl-os/ | GoHighLevel pipeline, contacts, invoices sync | 10-15 min |
ghl-github-bridge/ | Two-way sync between GHL deals and GitHub issues | 10-15 min |
calendly-os/ | Calendly call booking collection | 10-15 min |
seo-kit/ | SEO audit skill with checklists and schema patterns | 10-15 min |
To install: /install module-installs/content-pipeline (or any module path).
License
MIT — use, fork, and customise freely.