Orchestrator (/assistant)
The intelligent entry point that guides you to the right resources
What is the Orchestrator?
The orchestrator is the single entry point of claude-base. It analyzes your request, detects your project context, and points you to the most appropriate commands, agents and skills.
┌─────────────────────────────────────────────────────────────────┐
│ CLAUDE-BASE FOUNDATION │
├─────────────────────────────────────────────────────────────────┤
│ │
│ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ │
│ │ COMMANDS │ │ AGENTS │ │ SKILLS │ │
│ │ (123) │ │ (59) │ │ (42) │ │
│ │ │ │ │ │ │ │
│ │ Invocation │ │ Delegation │ │ Activation │ │
│ │ manual │ │ automatic │ │ automatic │ │
│ │ /xxx │ │ by Claude │ │ by context │ │
│ └─────────────┘ └─────────────┘ └─────────────┘ │
│ │
│ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ │
│ │ TEMPLATES │ │ RULES │ │ HOOKS │ │
│ │ (3) │ │ (24) │ │ (26) │ │
│ │ │ │ │ │ │ │
│ │ File │ │ Conventions │ │ Automation │ │
│ │ structures │ │ by path │ │ pre/post │ │
│ └─────────────┘ └─────────────┘ └─────────────┘ │
│ │
└─────────────────────────────────────────────────────────────────┘
Two available modes
| Command | Mode | Behavior |
|---|---|---|
/assistant | Guide | Analyze → Recommend → Wait for confirmation |
/assistant-auto | Automatic | Analyze → Execute the workflow directly |
Guide mode (/assistant)
For new users or when you want to validate before executing:
/assistant "Add an authentication feature"
# Claude analyzes and proposes:
# → Recommended workflow: /work:work-flow-feature
# → "Do you want me to launch this workflow?"
# → Waits for your confirmation
Automatic mode (/assistant-auto)
For advanced users who want immediate execution:
/assistant-auto "Add an authentication feature"
# Claude analyzes and executes directly:
# → Detects: new feature
# → Launches: /work:work-flow-feature "Add an authentication feature"
Automatic context detection
The orchestrator automatically detects your environment:
| Indicator | Project type | Recommended commands |
|---|---|---|
package.json + React/Next/Vue | Web Frontend | /dev:dev-component, /dev:dev-hook, /dev:dev-react-perf |
pubspec.yaml + Flutter | Mobile | /dev:dev-flutter, /dev:dev-supabase, /qa:qa-mobile |
package.json + Express/Fastify/NestJS | Node API | /dev:dev-api, /dev:dev-graphql, /dev:dev-trpc |
requirements.txt / pyproject.toml | Python | /dev:dev-api, /dev:dev-tdd |
go.mod | Go | /dev:dev-api, /dev:dev-tdd |
init.lua / .config/nvim | Neovim | /dev:dev-neovim, /qa:qa-neovim |
| Airflow/dbt/Spark | Data | /data:data-pipeline, /data:data-modeling |
Dockerfile / docker-compose.yml | DevOps | /ops:ops-docker, /ops:ops-k8s |
Proxmox / bpg/proxmox provider | Infrastructure | /ops:ops-proxmox, /ops:ops-infra-code |
Quick decision guide
┌────────────────────────────────────────────────────────────────────────┐
│ I WANT TO... → USE │
├────────────────────────────────────────────────────────────────────────┤
│ │
│ UNDERSTAND │
│ Explore a codebase → /work:work-explore │
│ Discover a new project → /doc:doc-onboard │
│ Understand complex code → /doc:doc-explain │
│ │
│ PLAN │
│ Create a specification → /work:work-specify │
│ Plan an implementation → /work:work-plan │
│ Define an MVP → /biz:biz-mvp │
│ │
│ DEVELOP │
│ Write code with tests → /dev:dev-tdd │
│ Create a React/Vue component → /dev:dev-component │
│ Create a REST API → /dev:dev-api │
│ Create a Flutter screen → /dev:dev-flutter │
│ Fix a bug → /dev:dev-debug │
│ │
│ VERIFY │
│ Code review → /qa:qa-review │
│ Security audit → /qa:qa-security │
│ Full audit → /qa:qa-audit │
│ │
│ DELIVER │
│ Create a commit → /work:work-commit │
│ Create a PR → /work:work-pr │
│ Publish a release → /ops:ops-release │
│ │
│ DEPLOY │
│ Dockerize → /ops:ops-docker │
│ Infrastructure as Code → /ops:ops-infra-code │
│ CI/CD → /ops:ops-ci │
│ │
└──────────────────────────────────────────── ────────────────────────────┘
Workflows by project type
Web (React/Next.js/Vue)
/work:work-explore → /work:work-specify → /work:work-plan → /dev:dev-component → /dev:dev-tdd → /qa:qa-review → /work:work-pr
Mobile (Flutter)
/work:work-explore → /work:work-specify → /work:work-plan → /dev:dev-flutter + /dev:dev-supabase → /qa:qa-mobile → /work:work-pr
Backend API (Node/Python/Go)
/work:work-explore → /work:work-specify → /work:work-plan → /dev:dev-api → /dev:dev-tdd → /qa:qa-security → /doc:doc-api-spec → /work:work-pr
Proxmox infrastructure
/work:work-explore → /ops:ops-proxmox → /ops:ops-monitoring → /ops:ops-backup
Sub-Agents activated automatically
The orchestrator knows the 63 specialized agents and activates them based on context:
| Detected context | Activated agent | Model |
|---|---|---|
| "Explore the code" | work-explore | haiku |
| "Security audit", "OWASP" | qa-security | sonnet |
| "Performance", "Core Web Vitals" | qa-perf | sonnet |
| "Accessibility", "WCAG" | wcag-audit | haiku |
| "Bug", "Debug" | dev-debug | sonnet |
| "Flutter", "Widget" | dev-flutter | sonnet |
| "Terraform", "IaC" | ops-infra-code | sonnet |
| "Proxmox", "VM", "LXC" | ops-proxmox | sonnet |
| "Docker", "Container" | ops-docker | haiku |
User: "Run a security audit"
│
▼
Claude detects: security → delegates to qa-security agent
│
▼
qa-security agent (isolated context, read-only)
│
▼
Result returned to the main conversation
Skills triggered automatically
The 54 skills activate based on keywords in the conversation:
| Keywords | Activated skill | Action |
|---|---|---|
| "TDD", "test first" | dev-tdd | Red-Green-Refactor cycle |
| "commit", "message" | work-commit | Conventional Commits |
| "review", "code review" | qa-review | In-depth review |
| "PR", "pull request" | work-pr | Structured PR |
| "Terraform", "IaC" | ops-infra-code | Infrastructure as Code |
| "Proxmox", "PVE" | ops-proxmox | Proxmox management |
| "Docker", "Dockerfile" | ops-docker | Containerization |
Decision flow
User: "/assistant I want to fix a login bug"
│
▼
┌─────────────────────────────────────┐
│ 1. REQUEST ANALYSIS │
│ - Keywords: "fix", "bug" │
│ - Domain: authentication │
└─────────────────────────────────────┘
│
▼
┌──────────────────── ─────────────────┐
│ 2. PROJECT DETECTION │
│ - package.json detected → Web │
│ - React detected → Frontend │
└─────────────────────────────────────┘
│
▼
┌─────────────────────────────────────┐
│ 3. RECOMMENDATION │
│ - Workflow: /work:work-flow-bugfix │
│ - Or manual steps: │
│ /work:work-explore → /dev:dev-debug │
│ → /dev:dev-test → /work:work-pr │
└─────────────────────────────────────┘
│
▼
┌─────────────────────────────────────┐
│ 4. PROPOSAL TO THE USER │
│ With explanations and options │
└─────────────────────────────────────┘
When to use the Orchestrator?
Use /assistant (guide mode) when:
- You are new to claude-base
- You don't know which command to use
- You want to validate the workflow before execution
- You are switching project type (Web → Mobile for example)
- You want an overview of the available options
Use /assistant-auto (automatic mode) when:
- You are familiar with claude-base
- You want immediate execution without confirmation
- You are doing repetitive tasks (features, bugfixes)
- You prefer speed over validation
Use direct commands when:
- You already know the exact command
- You want a fast, precise action
- You are in the middle of a workflow
Automatic mapping (/assistant-auto)
| Your request contains... | Executed workflow |
|---|---|
| feature, add, create | /work:work-flow-feature |
| bug, fix, error | /work:work-flow-bugfix |
| release, version, tag | /work:work-flow-release |
| launch, MVP, product | /work:work-flow-launch |
| security audit, OWASP | /qa:qa-security |
| full audit, quality | /qa:qa-audit |
| explore, understand | /work:work-explore |
| commit | /work:work-commit |
| PR, pull request | /work:work-pr |
| tests, TDD | /dev:dev-tdd |
| refactoring, clean up | /dev:dev-refactor |
| debug | /dev:dev-debug |
| Docker, container | /ops:ops-docker |
| CI/CD, pipeline | /ops:ops-ci |
Usage examples
New feature
/assistant "I want to add a push notification system"
# Orchestrator response:
# Project type: Mobile (Flutter detected)
# Recommended workflow:
# 1. /work:work-explore - Understand the current architecture
# 2. /work:work-specify - Specify the User Stories
# 3. /work:work-plan - Plan the implementation
# 4. /dev:dev-flutter - Create the widgets
# 5. /dev:dev-supabase - Configure the backend
# 6. /qa:qa-mobile - Test on devices
# 7. /work:work-pr - Create the PR
Bug fix
/assistant "The application crashes at login"
# Orchestrator response:
# Detected situation: Critical bug
# Recommended command: /work:work-flow-bugfix
# Or manual workflow:
# 1. /work:work-explore - Locate the problem
# 2. /dev:dev-debug - Investigate the cause
# 3. /dev:dev-test - Write a regression test
# 4. /work:work-commit - Commit the fix
General question
/assistant "How does authentication work in this project?"
# Orchestrator response:
# Request type: Exploration/Understanding
# Recommended command: /work:work-explore or /doc:doc-explain
# Automatically activated agent: work-explore (haiku)
Relationship with other concepts
/assistant (Orchestrator)
│
┌───────────────┼───────────────┐
│ │ │
▼ ▼ ▼
┌─────────┐ ┌─────────┐ ┌─────────┐
│COMMANDS │ │ AGENTS │ │ SKILLS │
│ │ │ │ │ │
│ Manual │ │ Auto │ │ Auto │
│ /xxx │ │ delegate│ │ keyword │
└─────────┘ └─────────┘ └─────────┘
│ │ │
└───────────────┴───────────────┘
│
┌──────┴──────┐
│ RULES │
│ HOOKS │
│ TEMPLATES │
└─────────────┘
The orchestrator is the conductor that:
- Understands your request
- Picks the right instruments (commands, agents, skills)
- Directs the workflow consistently
Best practices
- Start with
/assistantif you are new (guide mode) - Switch to
/assistant-autoonce familiar with the workflows - Be descriptive in your requests ("I want to..." rather than just "auth")
- Mention the context if relevant ("for the mobile app", "in production")
- Follow the proposed workflows for optimal results
See also
- Commands - Manual commands
- Agents - Autonomous sub-agents
- Skills - Auto-triggered skills
- Reference /assistant - Guide mode (with confirmation)
- Reference /assistant-auto - Automatic mode (direct execution)