Blueprints

The complete command and agent system for Claude Code. 40+ slash commands and 30+ specialized agents that give your AI assistant a structured methodology - from brainstorming to deployment, design to marketing.

$ npx @allthingsclaude/blueprints
Star us on GitHub↗
claude - ~/projects/openapi-sdk
> /brainstorm I want to build a CLI that converts OpenAPI specs into typed SDK clients Claude: Interesting - a few approaches here. We could do codegen via Handlebars templates, or an AST-based approach... > I like the AST approach. Target TS and Python first. > /critique Is AST-based codegen overkill for an MVP? Claude: For an MVP, yes - templates would ship 3x faster. I'd suggest we /plan this in phases. > /plan # Creates phased plan in tasks/plans/ > /bootstrap # Scaffolds project, installs deps, creates structure > /kickoff # Starts executing Phase 1 of the plan > /parallelize # Distributes Phase 2 tasks across subagents > /audit # Reviews all changes before committing > /test # Runs tests with analysis and fix suggestions > /finalize # Commits with proper message, updates plan docs
40+
Commands
Slash commands for every stage of development
30+
Agents
Specialized workers powered by tiered models
5
Power Levels
Balance capability with cost
n!
Workflows
Composable commands, infinite patterns

The Philosophy

Six principles that shape how every command and agent behaves.

01
Planning First
Understand before you build. Every task starts with research and planning, producing artifacts that guide implementation.
02
Structured Workflows
Commands compose into end-to-end pipelines. Each stage produces artifacts the next stage consumes.
03
Collaborative Execution
Choose your level of control. Interactive kickoff for complex work, autonomous implement for clear tasks, or full autopilot.
04
Continuous Validation
Built-in quality gates at every step. Audit, test, and security reviews before any code is committed.
05
Session Continuity
Context survives across sessions. Handoff serializes your state, pickup restores it. No lost work, no repeated explanations.
06
Convention Discovery
Agents scan your codebase for established patterns before implementing. They follow your conventions, not impose their own.

All Commands

Every command is a slash command you type directly in Claude Code. Each one launches a specialized workflow.

Planning & Execution
/planCreate structured implementation plans
/kickoffExecute plans with interactive approval gates
/implementExecute plans autonomously
/parallelizeDistribute tasks across multiple agents
/autopilotFull loop: idea to committed code
/finalizeWrap up, update plans, commit changes
Project Setup
/bootstrapGenerate plans and setup scripts for new projects
/onboardGenerate developer onboarding guides
Research & Learning
/researchSmart routing to codebase, docs, or web research
/explainDetailed code and architecture explanations
Code Quality
/auditPre-commit security and quality review
/testRun tests with analysis and fix suggestions
/cleanupRemove dead code and unused imports
/refactorSafe rename, extract, inline, move with validation
/dryEliminate DRY violations
/secureFocused security scanning
/a11yAudit frontend for accessibility issues
Code Operations
/commitCraft well-structured git commits
/changelogGenerate changelogs from git history
/docsGenerate or update documentation
/migrateUpgrade dependencies or framework versions
/mergeDetect parent branch and merge
/releaseVersion bump, changelog, tag, publish
/updateSync CLAUDE.md with project state
/diagramGenerate architecture diagrams
Thinking Modes
/brainstormPure ideation, no code changes
/challengeCritical analysis, question assumptions
/verifyQuick sanity check on your approach
/critiqueDirect, unfiltered feedback
Debugging
/debugSystematic investigation with root cause analysis
Creative & Design
/designCreate on-brand marketing assets for any platform
/copyWrite marketing copy for social media and ads
/brandCreate or document brand identity systems
/emailCreate HTML email templates
/pitchCreate presentation decks with speaker notes
/ogAuto-generate Open Graph images for all pages
/imagineGenerate images via AI
/storyboardExtract UI specs from video mockups
/showcaseDesign landing pages with animations
Session Management
/handoffGenerate context documentation for switching
/pickupResume work from a handoff document
/flushClear task artifacts
Utilities
/historyGit history narrative for files and functions
/i18nAudit and set up internationalization
/standupGenerate standup summary from git activity
/todoScan for TODO/FIXME/HACK markers

Agent System

Agents are specialized workers launched by commands. Each agent is assigned to a tier based on task complexity. Your chosen power level determines which Claude model runs each tier.

Lightweight
Rote tasks - fast models suffice
commit · changelog · handoff · cleanup · imagine · og
Research
Search and synthesize information
research-codebase · research-docs · research-web
Standard
Balanced reasoning and implementation
plan · implement · parallelize · bootstrap · refactor · test · explain · docs · dry · storyboard · migrate · onboard · diagram · i18n · release · showcase · design · copy · brand · email · pitch · update
Heavyweight
Deep reasoning, high-stakes analysis
audit · debug · secure · finalize · a11y
Power Levels

Control which models power your agents. Set during installation with --power 1-5.

Level
Lightweight
Research
Standard
Heavyweight
1. Economy
Haiku
Haiku
Haiku
Sonnet
2. Balanced
Haiku
Haiku
Sonnet
Sonnet
3. Standard
Sonnet
Sonnet
Sonnet
Sonnet
4. Enhanced*
Sonnet
Opus
Opus
Opus
5. Maximum
Opus
Opus
Opus
Opus

* Default power level. Higher levels use more capable (and expensive) models.

Lightweight Tier
commitGit commit message crafting
changelogChangelog generation
handoffContext documentation
cleanupDead code removal
imagineImage generation
ogOpen Graph image generation
Research Tier
research-codebaseCode exploration and search
research-docsLibrary documentation lookup
research-webOnline resource research
Standard Tier
planStructured plan creation
implementAutonomous plan execution
parallelizeMulti-agent orchestration
bootstrapProject scaffolding
refactorSafe code refactoring
testTest execution and analysis
explainCode explanations
docsDocumentation generation
dryDRY violation elimination
storyboardUI spec extraction
migrateDependency upgrades
onboardDeveloper onboarding
diagramArchitecture diagrams
i18nInternationalization
releaseRelease management
showcaseLanding page design
designMarketing asset creation
copyMarketing copywriting
brandBrand identity systems
emailHTML email templates
pitchPresentation deck creation
updateCLAUDE.md maintenance
Heavyweight Tier
auditCode quality and security
debugRoot cause investigation
secureVulnerability scanning
finalizeSession wrap-up
a11yAccessibility analysis

Workflows

Commands compose into end-to-end development patterns. Here are a few common examples - mix and match to build your own.

New Project
From idea to running project
/brainstorm→
/bootstrap→
/kickoff→
/test→
/finalize
Feature Development
Add features to existing projects
/research→
/plan→
/implement→
/audit→
/finalize
Bug Fix
Systematic debugging and verification
/debug→
/refactor→
/test→
/audit→
/finalize
Code Quality
Clean up and improve existing code
/audit→
/cleanup→
/dry→
/test→
/finalize
Marketing Campaign
From brand to published assets
/brand→
/design→
/copy→
/og→
/finalize
Session Handoff
Preserve context across sessions
/handoff→
/pickup
Full Automation
End-to-end autonomous development
/autopilot

Get Started

One command installs all 40+ slash commands and 30+ agents into your Claude Code setup.

$ npx @allthingsclaude/blueprints

No dependencies. No lock-in. Eject anytime - everything is plain Markdown files in your .claude directory.

View on GitHub↗
Installation Options
--globalInstall to ~/.claude (all projects)
--localInstall to current directory only
--path <dir>Install to a custom location
--power 1-5Set agent power level (default: 4)
--tasks-dir <name>Custom tasks directory name
--yesSkip confirmation prompts
Requirements
Node.js 16.0.0 or higher
Claude Code CLI
macOS, Linux, or Windows

Also from All Things Claude

Balcony
Coming Soon

Monitor Claude Code from your phone

Battery
→

Real-time API usage in your menu bar

Bar
→

Informative statusline for Claude Code