Skip to content

Latest commit

 

History

3 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

claude-stack

A unified Claude Code plugin that replaces 18 separate tools (GSD, Superpowers, Compound Engineering, gstack, ECC, wshobson, Uncodixfy, and more) with a single install. Ships with 27 skills, 12 agents, 6 hooks, and 8 rules for a complete development workflow -- from project initialization to shipping and knowledge capture.


Install

claude plugin add /path/to/claude-stack

# or when published:
# claude plugin install claude-stack

Quick Start

/s:new myproject     # Initialize project with .planning/ and docs/
/s:discuss           # Gather requirements interactively
/s:brainstorm        # Explore approaches, pick the best one
/s:plan              # Create phased roadmap with tasks
/s:build             # Implement with TDD (red/green/refactor)
/s:review            # Multi-perspective code review
/s:verify            # Evidence-based verification
/s:ship              # Test, push, and create PR
/s:compound          # Capture learnings for future sessions

All Skills

Skill Phase Description
/s:new Core Initialize project with .planning/ structure, docs/, CLAUDE.md, and git repo
/s:discuss Core Gather functional and non-functional requirements interactively
/s:plan Core Create phased implementation roadmap from requirements
/s:build Core Implement tasks using TDD (red/green/refactor cycle)
/s:status Core Show active phase, completed tasks, blockers, recent commits
/s:resume Core Reconstruct context from STATE.md and recent docs, suggest next action
/s:pause Core Save current work state and create resume instructions
/s:note Core Add timestamped note to STATE.md decisions table
/s:brainstorm Quality Explore 2-3 approaches with trade-offs, save to docs/brainstorms/
/s:review Quality 5-aspect code review: plan alignment, quality, architecture, security, docs
/s:verify Quality Build evidence table mapping requirements to test results
/s:debug Quality Systematic debugging: reproduce, root-cause, hypothesize, test, fix
/s:ceo-review Quality Product perspective review with scope analysis and prioritization
/s:eng-review Quality Technical review: architecture, data flow, edge cases, scalability
/s:ship Ship Sync, test, bump version, commit, push, and create PR
/s:compound Ship Extract and persist learnings to docs/solutions/ (never skip)
/s:retro Ship Retrospective: what went well, what didn't, action items
/s:docs Ship Detect stale documentation and update only what changed
/s:qa Browser Browser QA with 3 tiers: quick, standard, exhaustive
/s:qa-only Browser Report-only QA -- same analysis as /s:qa but no auto-fixes
/s:browse Browser Headless browser control: navigate, screenshot, fill forms, click
/s:cookies Browser Import cookies from installed browsers for authenticated testing
/s:auto Auto Run the full skill chain autonomously with circuit breaker
/s:quick Auto Compressed workflow for small tasks (discuss+plan+build in one go)
/s:do Auto Freeform natural language router to appropriate skill(s)
/s:map Auto Generate structured codebase documentation and dependency graph
/s:help Auto Show all skills, agents, and workflow reference card

All Agents

Agent Specialty When to Use
architect System design and trade-off analysis Major architectural decisions
code-reviewer 5-aspect review (plan, quality, arch, security, docs) Pre-merge code review
security-reviewer OWASP top 10, auth flows, injection, XSS Security-sensitive changes
database-reviewer Schema design, query performance, migration safety Database changes
tdd-guide RED/GREEN/REFACTOR coaching Learning or enforcing TDD
qa-engineer Test matrix, edge cases, coverage analysis QA planning and execution
frontend-dev React, Next.js, UI patterns Frontend implementation
backend-dev API design, service architecture Backend implementation
mobile-dev React Native, Swift patterns Mobile implementation
debugger Hypothesis-driven root cause analysis Hard-to-find bugs
performance-eng Profiling, bottleneck analysis Performance optimization
deployment-eng CI/CD, Docker, infrastructure Deploy and release

Additionally, 4 subagent templates are provided in agents/subagent-templates/ for implementer, spec-reviewer, code-quality-reviewer, and plan-reviewer roles.


Hooks

6 hooks fire automatically at key lifecycle events:

Hook Event What It Does
session-start SessionStart Loads STATE.md and recent git log to restore context
tdd-check PreToolUse (Write/Edit) Reminds you to write tests before implementation
post-edit PostToolUse (Write/Edit) Runs formatter and linter on saved files
pre-compact PreCompact Snapshots full project state before context compaction
statusline Notification Shows active phase and project status
stop Stop Reminds that state is preserved in .planning/ and docs/

Rules

8 rule sets are always active and enforced by the plugin's CLAUDE.md:

Rule What It Enforces
uncodixfy Anti-AI-slop: 33 "keep it normal" categories, 24 hard-no patterns
tdd No production code without a failing test first
verification No completion claims without fresh evidence (test output, screenshots)
debugging Systematic only -- no guessing, hypothesis-then-test required
context-persist Never delete .planning/ or docs/ directories
ask-user Re-ground, simplify, recommend, then present options
quality No any type, no emoji icons, use Lucide/Heroicons, code standards
confidence State confidence score (1-10) before major decisions

Workflow Chain

The recommended skill progression:

/s:new --> /s:discuss --> /s:brainstorm --> /s:plan --> /s:build --> /s:review --> /s:verify --> /s:ship --> /s:compound

Utility skills (use anytime):

  • /s:status -- check project state
  • /s:resume -- restore context in a new session
  • /s:pause -- save state before stepping away
  • /s:note -- quick capture to decisions log

Automation skills (shortcuts):

  • /s:auto -- run the full chain autonomously
  • /s:quick -- compressed workflow for small tasks
  • /s:do -- natural language router to any skill

Review add-ons (optional depth):

  • /s:ceo-review -- product perspective before building
  • /s:eng-review -- technical perspective before building

Browser skills (frontend projects):

  • /s:qa / /s:qa-only -- browser-based QA
  • /s:browse -- headless browser control
  • /s:cookies -- import auth cookies

Configuration

Context persistence: Skills read and write to .planning/ and docs/ directories in your project root. These directories are your project's memory across sessions.

Browser: The browser binary is not bundled. It is lazy-downloaded on first /s:qa or /s:browse use via bin/setup-browser.sh. If Playwright MCP is already configured, the plugin uses that instead.

MCP integrations: The following MCP servers work well with claude-stack but are configured separately:

  • Context7 -- pull library documentation before using a framework
  • Playwright -- browser automation for QA skills
  • Supabase -- database operations for backend workflows

Uninstall

claude plugin remove claude-stack

Note: .planning/ and docs/ directories are preserved -- they belong to your project, not the plugin. They are your project's memory.


License

MIT

About

Unified Claude Code plugin: 27 skills, 12 agents, 6 hooks, 8 rules. Merges GSD + Superpowers + gstack + Compound + Uncodixfy + ECC into one /s: namespace.

Resources

Stars

4 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages