原始内容
Wrangler
Wrangler is a Claude Code plugin that provides project governance, systematic workflows, and development tooling for AI-assisted software development. It ensures alignment between developers and AI agents through constitutional principles, tracked specifications, proven skill patterns, and automated quality enforcement.
What Wrangler Does
- Skills Library -- 40 proven workflow patterns covering TDD, debugging, code review, governance, and more
- MCP Server -- 18 tools for issue tracking, session orchestration, and workspace management
- Workflow Engine -- Automated spec-to-PR orchestration with phase-based execution and TDD enforcement
- Git Hooks -- Pre-commit and pre-push testing and quality gates
- Project Governance -- Constitution, roadmap, specifications, and issue tracking
Quick Start
Install Wrangler as a Claude Code plugin, then verify the installation:
/wrangler:help
Initialize the .wrangler/ workspace in your project:
/wrangler:init-workspace
Explore available commands and skills:
/wrangler:sitrep
/wrangler:issues
Skills Library
Skills are proven, reusable workflow patterns that Claude Code discovers and activates automatically when relevant to a task. When a skill exists for your current work, it is mandatory -- not optional.
40 skills organized across these areas:
| Category | Skills | Examples |
|---|---|---|
| Testing | 3 | TDD enforcement, test execution, anti-pattern detection |
| Debugging | 3 | Systematic debugging, root cause tracing, verification |
| Collaboration | 7 | Brainstorming, code review, implementation, parallel agents |
| Git Workflows | 6 | Worktrees, branch management, hooks setup |
| Issue Management | 3 | Issue creation, specifications, idea capture |
| Governance | 8 | Constitution, roadmap, metrics, alignment checks |
| Frontend | 4 | Design, E2E testing, visual regression, accessibility |
| Code Analysis | 5 | Code location, pattern finding, research |
| System | 3 | Wrangler usage, self-update, status reporting |
| Meta | 4 | Skill authoring, testing, sharing, file organization |
See docs/skill-invocation-patterns.md for task-to-skill mapping.
MCP Server
The built-in MCP server provides markdown-based tracking for issues, specifications, and workflow sessions. All data is stored as markdown files with YAML frontmatter in the .wrangler/ directory and version-controlled with git.
Issue Management (11 tools)
issues_create, issues_list, issues_search, issues_get, issues_update, issues_delete, issues_labels, issues_metadata, issues_projects, issues_mark_complete, issues_all_complete
Session Orchestration (6 tools)
session_start, session_phase, session_checkpoint, session_complete, session_get, session_status
Workspace Management (1 tool)
init_workspace
Issues and specifications are stored as {counter}-{slug}.md files (e.g., 000001-add-auth.md) with structured frontmatter for status, priority, labels, assignee, and project metadata.
See docs/mcp-usage.md for the complete tools reference.
Workflow Engine
The workflow engine (workflows/engine/) orchestrates complex spec-to-implementation workflows using the Claude Agent SDK. It provides deterministic, phase-based execution with built-in quality gates.
- Spec-implementation workflow -- Takes a specification and drives it through plan, execute, verify, and publish phases
- Code-review workflow -- Automated review with structured feedback
- Phase-based execution -- Each phase has defined entry/exit criteria
- Session tracking -- Checkpoints enable pause, resume, and recovery
- TDD enforcement -- Tests must pass before phase transitions
The engine is actively developed with 532 tests. See workflows/engine/ for details.
Slash Commands
14 commands that trigger specific workflows:
| Command | Purpose |
|---|---|
/wrangler:implement |
Implement from specs, plans, or issues |
/wrangler:run-tests |
Run tests and fix failures |
/wrangler:commit-push-pr |
Commit, push, and create PR |
/wrangler:write-plan |
Create an implementation plan |
/wrangler:generate-plan-for-spec |
Break a specification into issues |
/wrangler:issues |
Show issue and specification status |
/wrangler:idea |
Capture a new idea |
/wrangler:audit-session |
Audit session workflow adherence |
/wrangler:init-workspace |
Initialize .wrangler/ workspace |
/wrangler:sitrep |
Situational report on project state |
/wrangler:help |
Documentation and help |
/wrangler:setup-git-hooks |
Configure git hooks |
/wrangler:update-git-hooks |
Update hook configuration |
/wrangler:update-yourself |
Update wrangler plugin |
See docs/slash-commands.md for detailed usage.
Project Governance
Wrangler implements a three-tier governance framework:
- Constitution (
.wrangler/CONSTITUTION.md) -- Core design principles that guide all development decisions. Immutable unless formally amended. - Roadmap (
.wrangler/ROADMAP.md) -- Strategic planning with phased execution milestones. - Specifications and Issues -- Tactical execution. Specifications define planned work; issues track implementation tasks.
The governance system includes automated verification, constitutional alignment checks, and metric tracking. Run /wrangler:init-workspace to initialize governance files for a new project.
See docs/governance.md for the full governance guide.
Git Hooks
Automated testing and code quality enforcement through git hooks:
- Pre-commit -- Runs formatter, linter, and unit tests before each commit
- Pre-push -- Runs full test suite before pushing to protected branches
- Commit-msg -- Validates commit message format (optional)
TDD-aware: hooks can be bypassed during the RED phase when failing tests are expected. Run /wrangler:setup-git-hooks to configure.
See docs/git-hooks.md for details.
Directory Structure
wrangler/
├── skills/ # 40 workflow skills
├── commands/ # 14 slash commands
├── mcp/ # MCP server (TypeScript)
│ ├── tools/
│ │ ├── issues/ # 11 issue management tools
│ │ ├── session/ # 6 session orchestration tools
│ │ └── workspace/ # 1 workspace management tool
│ ├── providers/ # Storage providers (markdown)
│ ├── types/ # TypeScript types and schemas
│ └── __tests__/ # MCP test suite
├── workflows/
│ └── engine/ # Workflow engine (Claude Agent SDK)
│ ├── src/ # Engine source
│ └── __tests__/ # Engine test suite (532 tests)
├── hooks/ # Session hooks
├── docs/ # Documentation
├── .claude-plugin/ # Plugin configuration
└── .wrangler/ # Workspace (per-project)
├── CONSTITUTION.md # Design principles
├── ROADMAP.md # Strategic roadmap
├── issues/ # Issue tracking
├── specifications/ # Feature specifications
├── ideas/ # Ideas and proposals
├── plans/ # Implementation plans
├── memos/ # Reference material, RCAs
├── memory/ # Persistent agent memory
├── docs/ # Generated governance docs
├── config/ # Runtime configuration
├── orchestration/ # Workflow orchestration data
├── cache/ # Runtime cache (gitignored)
└── logs/ # Runtime logs (gitignored)
Documentation
| Document | Description |
|---|---|
| docs/mcp-usage.md | Complete MCP tools reference |
| docs/governance.md | Project governance framework |
| docs/workflows.md | Development workflows (TDD, verification, code review) |
| docs/slash-commands.md | Slash commands reference |
| docs/skill-invocation-patterns.md | Task-to-skill mapping |
| docs/session-hooks.md | Session hooks and state management |
| docs/git-hooks.md | Git hooks enforcement framework |
| docs/versioning.md | Versioning and updates |
| docs/verification-requirements.md | Evidence requirements for claims |
| docs/workflow-patterns.md | Multi-agent workflow patterns |
Development
MCP Server
npm run build:mcp # Build
npm run test:mcp # Run tests
npm run watch:mcp # Watch mode
Workflow Engine
cd workflows/engine
npm run build # Build
npm test # Run tests
npm run test:coverage # Coverage report
Requirements
- Claude Code with plugin support
- Git repository
- Node.js 20+
License
MIT