Aller au contenu principal

Cheatsheet - Claude Code Agents

Visual reference of all available agents.

╔═══════════════════════════════════════════════════════════════════════════════╗
║ CLAUDE CODE AGENTS - CHEATSHEET ║
╠═══════════════════════════════════════════════════════════════════════════════╣
║ Total: 106 commands | 45 agents | 53 skills | 9 categories ║
╚═══════════════════════════════════════════════════════════════════════════════╝

┌─────────────────────────────────────────────────────────────────────────────┐
│ ASSISTANT │
├─────────────────────────────────────────────────────────────────────────────┤
│ ┌────────────────┐ ┌────────────────────┐ │
│ │ /assistant │ │ /assistant-auto │ │
│ │ Guide mode │ │ Automatic mode │ │
│ └────────────────┘ └────────────────────┘ │
│ Orchestrators: analyze intent and route to the right workflow. │
└─────────────────────────────────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────────────────────────────────┐
│ WORK- : MAIN WORKFLOW │
├─────────────────────────────────────────────────────────────────────────────┤
│ │
│ EXPLORE → (BRAINSTORM) → SPECIFY → PLAN → TDD → AUDIT → COMMIT │
│ │
│ ┌─────────┐ ┌──────────┐ ┌─────────┐ ┌──────┐ ┌─────┐ ┌───────┐ ┌────────┐ │
│ │ EXPLORE │→│BRAINSTORM│→│ SPECIFY │→│ PLAN │→│ TDD │→│ AUDIT │→│ COMMIT │ │
│ │ Read │ │ (option) │ │ Stories │ │ Plan │ │ R-G-R│ │ Score │ │ +PR │ │
│ │ existing│ │ Ideation │ │ Criteria│ │ Files│ │ Tests│ │ ≥ 90 │ │ │ │
│ └─────────┘ └──────────┘ └─────────┘ └──────┘ └─────┘ └───────┘ └────────┘ │
│ │
│ Commands: /work:work-explore, /work:work-brainstorm, │
│ /work:work-specify, /work:work-plan, /dev:dev-tdd, │
│ /qa:qa-loop "score 90", /work:work-commit, /work:work-pr │
│ │
│ Chained workflows (all-in-one): │
│ ┌────────────────┐ ┌────────────────┐ ┌────────────────┐ ┌─────────────┐ │
│ │ flow-feature │ │ flow-bugfix │ │ flow-release │ │ flow-launch │ │
│ │ Feature A→Z │ │ Bug A→Z │ │ Release A→Z │ │ Product A→Z │ │
│ └────────────────┘ └────────────────┘ └────────────────┘ └─────────────┘ │
│ │
│ Variants: │
│ ┌──────────┐ ┌──────────┐ ┌──────────┐ │
│ │ quick │ │ batch │ │ team │ │
│ │ <50 LOC │ │ Multi-US │ │ Agents │ │
│ └──────────┘ └──────────┘ └──────────┘ │
└─────────────────────────────────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────────────────────────────────┐
│ DEV- : DEVELOPMENT (16 commands) │
├─────────────────────────────────────────────────────────────────────────────┤
│ TDD cycle : tdd (cycle + test-gen + infra setup), debug, refactor │
│ API & types : api (REST/GraphQL/tRPC + versioning), prisma │
│ UI & front : component (+hooks), design-system, react-perf, error-handling│
│ Mobile/Edit : flutter, neovim │
│ AI / Document : ai-integration, rag, mcp, document │
│ Backend BaaS : supabase │
│ → Full details: "DEV- : Development (16)" table below │
└─────────────────────────────────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────────────────────────────────┐
│ QA- : QUALITY (13 commands) │
├─────────────────────────────────────────────────────────────────────────────┤
│ Main audit : qa-loop (audit + fix loop, score ≥ 90) │
│ Targeted audits : security, perf, design (UI/UX + responsive), tech-debt │
│ (debt + coverage + kaizen), review, audit, automation,e2e│
│ Specific : wcag-audit (a11y), mobile, neovim, chrome (visual) │
│ → Full details: "QA- : Quality (13)" table below │
└─────────────────────────────────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────────────────────────────────┐
│ OPS- : OPERATIONS (28 commands) │
├─────────────────────────────────────────────────────────────────────────────┤
│ Release & deploy : release, deploy, rollback, hotfix, mobile-release │
│ GitFlow : gitflow <init|feature|release|hotfix> <action> │
│ CI/CD : ci, ci-fix, deps │
│ Infra : docker, k8s, vps, vercel, serverless, │
│ proxmox, opnsense, infra-code │
│ Observability : monitoring (+stack deploy), grafana-dashboard, │
│ load-testing, health │
│ Data/Sec : database, backup (+DR), secrets-management │
│ Meta-ops : env, migrate, cost (tokens+cloud FinOps), standup │
│ → Full details: "OPS- : Operations (28)" table below │
└─────────────────────────────────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────────────────────────────────┐
│ DOC- : DOCUMENTATION │
├─────────────────────────────────────────────────────────────────────────────┤
│ │
│ ┌──────────┐ ┌───────────┐ ┌─────────┐ ┌─────────┐ ┌──────────┐ │
│ │ GENERATE │ │ CHANGELOG │ │ EXPLAIN │ │ ONBOARD │ │ API-SPEC │ │
│ │ Auto doc │ │ History │ │ Teach │ │ Discover│ │ OpenAPI │ │
│ └──────────┘ └───────────┘ └─────────┘ └─────────┘ └──────────┘ │
│ (generate covers README + architecture/ADR doc types) │
└─────────────────────────────────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────────────────────────────────┐
│ BIZ- : BUSINESS │
├─────────────────────────────────────────────────────────────────────────────┤
│ │
│ ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌─────────┐ │
│ │ MODEL │ │ MVP │ │ PRICING │ │ PITCH │ │
│ │ Canvas │ │ Minimum │ │ Strategy│ │ Deck │ │
│ └─────────┘ └─────────┘ └─────────┘ └─────────┘ │
│ │
│ ┌──────────────┐ ┌─────────┐ ┌──────────────────┐ ┌──────────┐ │
│ │ ROADMAP +OKR │ │ LAUNCH │ │ COMPETITOR +market│ │ RESEARCH │ │
│ │ Plan + OKRs │ │ Go-to │ │ TAM/SAM + SWOT │ │ User res │ │
│ └──────────────┘ └─────────┘ └──────────────────┘ └──────────┘ │
│ │
│ ┌──────────┐ │
│ │ PERSONAS │ │
│ │ Profiles │ │
│ └──────────┘ │
└─────────────────────────────────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────────────────────────────────┐
│ GROWTH- : GROWTH (9 commands) │
├─────────────────────────────────────────────────────────────────────────────┤
│ Acquisition : landing, seo, ab-test │
│ Conversion : cro (funnel + onboarding/activation) │
│ Measurement : analytics, app-store-analytics │
│ Engagement : email, retention │
│ International : localization │
│ → Full details: "GROWTH- : Growth (9)" table below │
└─────────────────────────────────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────────────────────────────────┐
│ DATA- : DATA │
├─────────────────────────────────────────────────────────────────────────────┤
│ │
│ ┌──────────┐ ┌──────────┐ │
│ │ PIPELINE │ │ MODELING │ │
│ │ ETL/ELT │ │ DWH │ │
│ └──────────┘ └──────────┘ │
└─────────────────────────────────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────────────────────────────────┐
│ LEGAL- : LEGAL │
├─────────────────────────────────────────────────────────────────────────────┤
│ │
│ ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌──────────────────┐ │
│ │ DOCS │ │ GDPR │ │ PAYMENT │ │ TERMS-OF-SERVICE │ │
│ │ T&Cs │ │ GDPR │ │ Stripe │ │ Detailed ToS │ │
│ └─────────┘ └─────────┘ └─────────┘ └──────────────────┘ │
│ │
│ ┌────────────────┐ │
│ │ PRIVACY-POLICY │ │
│ │ Privacy │ │
│ └────────────────┘ │
└─────────────────────────────────────────────────────────────────────────────┘

╔═══════════════════════════════════════════════════════════════════════════════╗
║ QUICK REFERENCE ║
╠═══════════════════════════════════════════════════════════════════════════════╣
║ ║
║ MANDATORY WORKFLOW: ║
║ ┌────────────────────────────────────────────────────────────────────────┐ ║
║ │ EXPLORE → (BRAINSTORM) → SPECIFY → PLAN → TDD → AUDIT → COMMIT │ ║
║ │ explore brainstorm specify plan dev-tdd qa-loop commit │ ║
║ └────────────────────────────────────────────────────────────────────────┘ ║
║ ║
║ NEW FEATURE: /work:work-flow-feature "description" ║
║ BUG FIX: /work:work-flow-bugfix "issue #123" ║
║ NEW RELEASE: /work:work-flow-release "v2.0" ║
║ PRODUCT LAUNCH: /work:work-flow-launch "my SaaS" ║
║ ║
║ FULL AUDIT: /qa:qa-audit ║
║ QUICK HEALTH CHECK: /ops:ops-health ║
║ ║
╚═══════════════════════════════════════════════════════════════════════════════╝

Commands by Category (106)

Orchestrator (1)

CommandUsage
/assistantGuide for choosing agents and workflows

WORK- : Main Workflow (15)

CommandUsage
/work:work-exploreExplore and understand the code
/work:work-brainstormStructured ideation before spec
/work:work-specifyCreate a functional specification
/work:work-clarifyClarify ambiguities
/work:work-planPlan an implementation
/work:work-commitCreate a clean commit
/work:work-prCreate a Pull Request
/work:work-commit-push-prCommit + push + PR in a single command
/work:work-quickQuick workflow (trivial changes, skip full cycle)
/work:work-batchSequential execution of user stories from a PRD
/work:work-teamLaunch a team of coordinated agents (Agent Teams)
/work:work-flow-featureFull feature workflow
/work:work-flow-bugfixFull bugfix workflow
/work:work-flow-releaseFull release workflow
/work:work-flow-launchFull launch workflow

DEV- : Development (16)

CommandUsage
/dev:dev-tddTDD cycle + test generation + test infra setup
/dev:dev-debugDebug a problem
/dev:dev-refactorGuided refactoring
/dev:dev-documentDocument generation (PDF, DOCX, XLSX, PPTX)
/dev:dev-apiCreate/document REST, GraphQL, or tRPC API + versioning
/dev:dev-componentCreate a complete UI component (or custom hook)
/dev:dev-error-handlingError handling strategy
/dev:dev-react-perfReact/Next.js optimization
/dev:dev-mcpCreate MCP servers
/dev:dev-flutterFlutter widgets and screens
/dev:dev-supabaseSupabase backend (Auth, DB, Storage)
/dev:dev-neovimNeovim plugins and config
/dev:dev-design-systemDesign tokens and components
/dev:dev-prismaPrisma ORM
/dev:dev-ragRAG systems
/dev:dev-ai-integrationLLM integration (OpenAI, Claude)

QA- : Quality (13)

CommandUsage
/qa:qa-reviewCode review + naming analysis
/qa:qa-securityOWASP security audit
/qa:qa-perfPerformance analysis
/qa:wcag-auditWCAG accessibility audit
/qa:qa-auditFull audit (all in one)
/qa:qa-designUI/UX audit (100+ rules, incl. responsive/mobile)
/qa:qa-automationTest automation
/qa:qa-e2eE2E tests (Playwright, Cypress)
/qa:qa-mobileMobile app quality audit
/qa:qa-neovimNeovim config audit
/qa:qa-tech-debtTechnical debt (incl. coverage + kaizen)
/qa:qa-chromeChrome visual tests
/qa:qa-loopAudit + fix loop until target score (90 by default)

OPS- : Operations (28)

CommandUsage
/ops:ops-hotfixUrgent production fix
/ops:ops-releaseCreate a release
/ops:ops-rollbackSafe rollback
/ops:ops-gitflowGitFlow: <init|feature|release|hotfix> <action>
/ops:ops-depsDependency audit and updates
/ops:ops-dockerDockerize
/ops:ops-k8sKubernetes deployment
/ops:ops-vpsVPS deployment
/ops:ops-migrateCode/deps migration
/ops:ops-ciCI/CD pipelines
/ops:ops-monitoringLogs, metrics, alerts + deploy stack (Prometheus/Grafana/Loki)
/ops:ops-grafana-dashboardGrafana dashboards
/ops:ops-databaseDB schema, migrations
/ops:ops-healthQuick health check
/ops:ops-envEnvironment management
/ops:ops-backupBackup/restore + disaster recovery (RPO/RTO)
/ops:ops-load-testingLoad testing
/ops:ops-infra-codeInfrastructure as Code
/ops:ops-proxmoxProxmox VE infrastructure
/ops:ops-opnsenseOPNsense configuration
/ops:ops-secrets-managementSecrets management
/ops:ops-serverlessServerless deployment
/ops:ops-vercelVercel configuration
/ops:ops-mobile-releaseApp/Play Store publishing
/ops:ops-deploySafe deployment with pre-deploy checklist
/ops:ops-costClaude Code token tracking and costs
/ops:ops-standupCross-repo morning briefing
/ops:ops-ci-fixAutonomous CI diagnostic and repair

DOC- : Documentation (5)

CommandUsage
/doc:doc-generateGenerate documentation (README, architecture/ADR, API, inline)
/doc:doc-changelogChangelog
/doc:doc-explainExplain complex code
/doc:doc-onboardDiscover a codebase
/doc:doc-api-specOpenAPI/Swagger spec

BIZ- : Business (9)

CommandUsage
/biz:biz-modelBusiness model, Lean Canvas
/biz:biz-mvpDefine the MVP
/biz:biz-pricingPricing strategy
/biz:biz-pitchPitch deck
/biz:biz-roadmapProduct roadmap + OKRs
/biz:biz-launchLaunch workflow
/biz:biz-competitorMarket study + competitive analysis
/biz:biz-personasUser personas
/biz:biz-researchUser research

GROWTH- : Growth (9)

CommandUsage
/growth:growth-landingLanding page
/growth:growth-seoSEO audit
/growth:growth-analyticsTracking and KPIs
/growth:growth-app-store-analyticsApp/Play Store metrics
/growth:growth-emailEmail templates
/growth:growth-ab-testA/B testing
/growth:growth-retentionRetention strategies
/growth:growth-localizationMulti-market localization
/growth:growth-croCRO + funnel + onboarding/activation

DATA- : Data (2)

CommandUsage
/data:data-pipelineETL/ELT pipelines
/data:data-modelingData warehouse modeling
CommandUsage
/legal:legal-docsT&Cs, Sales T&Cs, legal notices
/legal:legal-rgpdGDPR compliance
/legal:legal-paymentPayment integration
/legal:legal-terms-of-serviceTerms of Service
/legal:legal-privacy-policyPrivacy Policy

Quick decision by intent

"I want to do X" → which command to use?

IntentCommand
Understand existing code/work:work-explore
Brainstorm / explore ideas/work:work-brainstorm
Create a specification (User Stories)/work:work-specify
Clarify ambiguities in a spec/work:work-clarify
Plan a feature (architecture, tasks)/work:work-plan
Implement with TDD/dev:dev-tdd
Debug a problem/dev:dev-debug
Refactor code/dev:dev-refactor
Do a code review/qa:qa-review
Full audit + fix loop/qa:qa-loop
Check security (OWASP)/qa:qa-security
Audit performance/qa:qa-perf
Check accessibility (WCAG)/qa:wcag-audit
Full audit (read-only)/qa:qa-audit
Create a clean commit/work:work-commit
Create a PR/work:work-pr
Morning briefing / standup/ops:ops-standup
Fix broken CI/ops:ops-ci-fix
Deploy to production/ops:ops-deploy
Quick health check/ops:ops-health

If you hesitate: /assistant (guide mode) or /assistant-auto "your request" (direct execution).


claude-base CLI (install & modules)

claude-base init [--preset <name>] . Install (core-only by default; auto-detects stack)
claude-base modules List modules + what's installed
claude-base add <module> . Opt into a module (biz, growth, legal, mobile, iac…)
claude-base remove <module> . Opt back out
claude-base update [--all] . Update; keeps the preset filter + flags recommendation/stack drift
claude-base update --detect-only . Read-only: does the project still match its recorded preset?
claude-base validate . Check the install

Default install is core-only (work, dev, qa, ops, doc). Horizontal domains, the data domain (data-eng module) and platform/stack tooling are opt-in modules.

Common Scenarios

New project

/doc:doc-onboard → Understand the structure
/work:work-explore → Explore the code
/work:work-plan → Plan the work

New feature (standard workflow)

/work:work-explore → Understand the existing code
/work:work-specify → Spec User Stories + Given/When/Then criteria
/work:work-plan → Design the solution + tasks per US
/dev:dev-tdd → Implement with tests (Red-Green-Refactor)
/qa:qa-loop "score 90" → Audit + fix loop (mandatory before commit)
/work:work-commit → Commit cleanly
/work:work-pr → Create the PR

Trivial change (skip full cycle)

/work:work-quick "rename X to Y" # < 50 LOC, 1-3 files

New feature (full workflow)

/work:work-flow-feature "add dark mode"

Bug fix

/work:work-flow-bugfix "#123 - user cannot log in"

Before going to production

/qa:qa-audit → Full audit
/ops:ops-health → Quick health check

New release

/work:work-flow-release "v2.0.0"

Launch a new business

/work:work-flow-launch "my new SaaS"

Flutter mobile app

/work:work-explore → /work:work-plan → /dev:dev-flutter + /dev:dev-supabase → /qa:qa-mobile → /work:work-pr

Command Format

/{category}-{action} "context"

Categories:
• assistant → Orchestrator
• work- → Base workflow
• dev- → Development
• qa- → Quality
• ops- → Operations
• doc- → Documentation
• biz- → Business
• growth- → Growth
• data- → Data
• legal- → Legal

Built-in Claude Code Commands

CommandUsage
/clearReset the context
/compactCompact the history
/resumeResume a session
/helpHelp

Keyboard Shortcuts

ShortcutAction
EscapeInterrupt Claude
Escape x2Go back
TabAutocompletion
Ctrl+CCancel

Reasoning Keywords

KeywordLevel
thinkBasic
think hardDeep
think harderDeeper
ultrathinkMaximum

Commit Format

type(scope): description

Types: feat, fix, refactor, test, docs, style, chore, perf, hotfix

Examples:

feat(auth): add OAuth2 login
fix(api): handle null response
refactor(user): extract validation logic

Configuration Files

FileRole
CLAUDE.mdProject instructions
CLAUDE.local.mdLocal config (gitignore)
.claude/settings.jsonPermissions and hooks
.claude/commands/**/*.mdCommands (organized by category)
.claude/skills/Automatic skills
.claude/agents/Isolated sub-agents
.claude/rules/Path-based contextual rules
.mcp.jsonMCP servers

Golden Rules

  1. Explore before coding - Always understand the context
  2. Plan before implementing - Avoids backtracking
  3. Tests BEFORE code (TDD) - Mandatory Red-Green-Refactor cycle
  4. Audit before commit - qa-loop score ≥ 90 mandatory before push
  5. Atomic commits - One commit = one concern
  6. Conventional Commits - feat(scope): description
  7. No secrets in code - Use .env + gitleaks

Resources


Claude-Base v<!-- version -->5.1.0<!-- /version --> - 106 commands - 44 agents - 53 skills - 32 rules