A collection of powerful plugins for Claude Code, featuring automated development workflows and multi-perspective decision-making systems.
| Plugin | Description | Command |
|---|---|---|
| Dev Workflow | Complete development workflow from requirements to QA | /dev-workflow |
| Multi-Agent Debate | Multi-perspective analysis with critical review | /debate |
| High-Precision Dev | Safety-critical code with p^4 error rate compression | /init, /start |
# Add the marketplace
/plugin marketplace add chinlung/claude-dev-workflow
# Install all plugins
/plugin install dev-workflow@scl-claude-plugins
/plugin install multi-agent-debate@scl-claude-plugins
/plugin install high-precision-dev@scl-claude-pluginsOr install directly:
/plugin install chinlung/claude-dev-workflowA comprehensive development workflow system for Claude Code, automating the journey from requirements analysis to quality assurance.
Inspired by the 8-role workflow system from Kiro IDE, designed by Pahud Hsieh. Watch the tutorial video to learn more about the original concept.
- 7 Specialized Agents working in sequence
- Structured Documentation output for every phase
- handoff.md Mechanism for seamless context transfer between agents
- Pause Point after architecture design for user confirmation
- Progress Tracking with TodoWrite
- Flexible Execution: Full workflow, single step, or resume from checkpoint
- Language Agnostic: Works with any programming language
/dev-workflow Implement user authentication featureThis executes all 7 phases in sequence:
- Requirements Analysis →
01-requirements-analysis.md - Code Exploration →
02-code-analysis.md - Architecture Design →
03-architecture-design.md(pauses for confirmation) - Implementation →
04-implementation-report.md - Testing →
05-test-report.md - Quality Assurance →
06-quality-report.md - Documentation →
07-documentation-report.md
/dev-workflow --step analyze Analyze shopping cart requirements
/dev-workflow --step explore Explore authentication code
/dev-workflow --step design Design payment integration
/dev-workflow --step implement Implement the feature
/dev-workflow --step test Write tests for the feature
/dev-workflow --step qa Run quality checks
/dev-workflow --step docs Update documentation and generate PR/dev-workflow --resume docs/task-20241211-1430-auth-feature/| Agent | Role | Output |
|---|---|---|
| Issue Analyst | Requirements analysis, user stories, success criteria | 01-requirements-analysis.md |
| Code Archaeologist | Codebase exploration, pattern identification, reusable components | 02-code-analysis.md |
| Solution Architect | Architecture design, solution comparison, implementation plan | 03-architecture-design.md |
| Implementation Specialist | Code implementation following best practices | 04-implementation-report.md |
| Test Engineer | Test planning, writing, and execution | 05-test-report.md |
| Quality Assurance | Lint, type check, build verification, code review | 06-quality-report.md |
| Documentation Specialist | README, CHANGELOG, API docs, PR description | 07-documentation-report.md |
All reports are saved in a task directory:
docs/task-{YYYYMMDD-HHMM}-{brief-name}/
├── 01-requirements-analysis.md
├── 02-code-analysis.md
├── 03-architecture-design.md
├── 04-implementation-report.md
├── 05-test-report.md
├── 06-quality-report.md
├── 07-documentation-report.md
├── handoff.md
└── summary.md
┌─────────────────┐
│ Issue Analyst │ → Requirements & User Stories
└────────┬────────┘
▼
┌─────────────────┐
│Code Archaeologist│ → Codebase Analysis
└────────┬────────┘
▼
┌─────────────────┐
│Solution Architect│ → Architecture Design
└────────┬────────┘
▼
⏸️ PAUSE (User Confirmation)
▼
┌─────────────────┐
│Implementation │ → Working Code
│ Specialist │
└────────┬────────┘
▼
┌─────────────────┐
│ Test Engineer │ → Test Suite
└────────┬────────┘
▼
┌─────────────────┐
│Quality Assurance│ → Quality Report
└────────┬────────┘
▼
┌─────────────────┐
│ Documentation │ → Docs & PR Description
│ Specialist │
└─────────────────┘
A multi-agent dialectical system that generates optimal solutions through multi-perspective analysis and critical review.
- 5 Specialized Agents working collaboratively
- Smart Perspective Configuration based on requirement type
- Quantitative Scoring System (30-point scale)
- Consensus-Driven Decision Making (≥2 agents must agree)
- Iterative Refinement through multiple debate rounds
- User Participation at key decision points
/debate Design a caching strategy for the API layer/debate <requirement> [--max-rounds N] [--perspectives "angle1,angle2,angle3"]Examples:
/debate Should we use microservices or monolith for the new project?
/debate --max-rounds 5 How to optimize database query performance?
/debate --perspectives "security,performance,maintainability" Design the authentication system| Agent | Role |
|---|---|
| Orchestrator | Analyzes requirements, configures perspectives, manages workflow |
| Perspective A | Proposes solutions from assigned angle |
| Perspective B | Proposes solutions from assigned angle |
| Perspective C | Proposes solutions from assigned angle |
| Critic | Reviews all proposals, raises challenges, provides quantitative scoring |
┌─────────────────┐
│ Orchestrator │ → Analyze & Configure Perspectives
└────────┬────────┘
▼
┌─────────────────────────────────────┐
│ Perspective A │ B │ C (Parallel) │ → Initial Proposals
└────────┬────────────────────────────┘
▼
┌─────────────────┐
│ Critic │ → Review & Challenge
└────────┬────────┘
▼
┌─────────────────────────────────────┐
│ Perspective A │ B │ C (Parallel) │ → Rebuttals & Revisions
└────────┬────────────────────────────┘
▼
┌─────────────────┐
│ Consensus Check │ → ≥2 agree? → Done
└────────┬────────┘ Otherwise ↩️ Back to Critic
▼
┌─────────────────┐
│ User Decision │ → Continue / Adopt / Intervene / Reset
└─────────────────┘
The Orchestrator automatically selects perspectives based on requirement type:
| Requirement Type | Perspective A | Perspective B | Perspective C |
|---|---|---|---|
| Architecture Design | Performance-First | Maintainability-First | Scalability-First |
| Feature Development | Fast Delivery | Quality-First | UX-First |
| Performance Optimization | Algorithm Optimization | Caching Strategy | Architecture Refactoring |
| Bug Fixing | Quick Patch | Root Cause Fix | Defensive Refactoring |
| Technology Selection | Mainstream & Stable | Emerging Tech | Custom Solution |
| Security Issues | Least Privilege | Defense in Depth | Zero Trust |
| User Experience | Simplify Flow | Add Guidance | Customization |
| Cost Control | Minimum Cost | Balanced Approach | Long-term Investment |
| Refactoring | Incremental | Complete Rewrite | Hybrid Strategy |
The Critic evaluates each proposal on three dimensions (10 points each, 30 total):
| Dimension | Criteria |
|---|---|
| Feasibility | Technical achievability, resource availability, timeline reasonability |
| Benefit | Problem resolution degree, positive value, ROI |
| Risk Control | Risk identification completeness, mitigation reliability, failure impact scope |
Score Guide:
- 9-10: Excellent - Complete solution / Mature technology
- 7-8: Good - Mostly solved / Minor preparation needed
- 5-6: Fair - Partially solved / Challenges exist but manageable
- 3-4: Poor - Limited benefit / Significant resources required
- 1-2: Very Poor - Minimal benefit / Questionable feasibility
- Architecture design decisions
- Feature development planning
- Performance optimization strategies
- Bug fix approaches
- Technology stack selection
- Security solution design
- UX improvement strategies
- Cost optimization plans
- Refactoring task planning
A multi-agent development mode that compresses single-agent error rate from p to p^4 through epistemic division of labor. Designed for safety-critical code: cryptography, financial calculations, data validation, and security-critical logic.
This plugin is the highest level of a three-level intensity spectrum:
Single Agent (~1x) → /debate (~4-5x) → /start (~12-18x)
General CRUD Decisions Safety-critical code
Chain them: Use /debate to decide What/Why, then /start to implement How.
/init parse_amountThis creates SPEC.md and CONSENSUS.md templates. Fill in SPEC.md completely (especially the boundary conditions table).
/start ./SPEC.mdThis runs the full 4-phase workflow:
- Spec Review — All agents review SPEC.md for ambiguity
- Dual Implementation — Two implementers work independently in isolated worktrees
- Adversarial Review — Critic finds bugs + Adversary attacks with 3 rounds
- Integration — Verifier merges the best parts with full SPEC.md coverage
| Agent | Role | Output |
|---|---|---|
| Implementer A | Independent defensive implementation | IMPL_A_REPORT.md |
| Implementer B | Independent defensive implementation | IMPL_B_REPORT.md |
| Critic | Find bugs using severity 1-5 scale | CRITIQUE.md |
| Adversary | 3-round red team attack (boundary, semantic, assumption) | ATTACKS.md |
| Verifier | Compare, merge, verify against SPEC.md | VERIFICATION.md |
┌─────────────────────────────────────┐
│ All Agents (except Verifier) │ → Spec Review
└────────────────┬────────────────────┘
▼
┌─────────────────────────────────────┐
│ Implementer A │ B (Parallel) │ → Independent Implementation
│ (Worktree Isolation) │
└────────────────┬────────────────────┘
▼
┌─────────────────────────────────────┐
│ Critic │ Adversary (Parallel) │ → Adversarial Review
└────────────────┬────────────────────┘
severity ≥ 3 │ ↩️ Fix cycle (max 3)
▼
┌─────────────────────────────────────┐
│ Verifier │ → Final Integration
└─────────────────────────────────────┘
| Use This | Use /debate Instead |
Use Single Agent |
|---|---|---|
| Core algorithms | Architecture decisions | General CRUD |
| Cryptography | Technology selection | UI adjustments |
| Financial calculations | Design trade-offs | Config changes |
| Data validation | Refactoring strategy | Quick prototypes |
| Security-critical logic | Spec validation |
All plugins work out of the box with any project. Each agent adapts to your project's structure and conventions.
For project-specific coding standards, consider creating a CLAUDE.md file in your project root with your team's guidelines.
MIT
Contributions are welcome! Please feel free to submit a Pull Request.
- Dev Workflow inspired by the 8-role workflow system from Kiro IDE, designed by Pahud Hsieh
- Tutorial video explaining the original workflow concept
- Built for Claude Code