Show HN: Claude Code plugin that runs a semi-autonomous AI dev workflow

1 month ago 6

Professional development workflows for Claude Code - Language-agnostic best practices, specialized agents, and quality assurance patterns for building production-ready software.

# 1. Start Claude Code session claude # 2. Inside Claude Code, install the plugin /plugin marketplace add shinpr/claude-code-workflows /plugin install claude-code-workflows@shinpr # 3. Start building with full workflow support /implement <your feature>

Note: If you encounter SSH authentication errors, set up SSH keys for GitHub:

# 1. Check if SSH key already exists ls ~/.ssh/id_ed25519.pub # 2. Generate new SSH key (if needed) ssh-keygen -t ed25519 -C "[email protected]" # → Press Enter to save to default location # → Enter a strong passphrase when prompted (recommended for security) # 3. Add SSH key to ssh-agent eval "$(ssh-agent -s)" ssh-add ~/.ssh/id_ed25519 # 4. Copy public key to clipboard cat ~/.ssh/id_ed25519.pub # → Copy the output # 5. Add to GitHub # Go to https://github.com/settings/keys # Click "New SSH key" # Paste your public key and save # 6. Test connection ssh -T [email protected] # → Should see: "Hi username! You've successfully authenticated..."

11 production-ready agents for every phase of development:

Agent Purpose Activation

requirement-analyzer	Assess task scope and complexity	Start of /implement
prd-creator	Create product requirements documents	Large-scale features
technical-designer	Design architecture and technical approach	Medium/large features
work-planner	Break down work into executable tasks	After design
task-decomposer	Decompose plans into commit-sized tasks	Task breakdown
task-executor	Execute individual tasks with TDD	Implementation
quality-fixer	Fix all quality issues automatically	After code changes
code-reviewer	Verify compliance with design docs	Post-implementation
acceptance-test-generator	Generate E2E and integration tests	Testing phase
rule-advisor	Select optimal rules for current task	Task initiation
document-reviewer	Review documentation consistency	Documentation phase

Streamline your development with purpose-built commands:

Command Purpose When to Use

/implement	End-to-end feature development	New features, complete workflows
/task	Execute single task with precision	Bug fixes, small changes
/design	Create design documentation	Architecture planning
/plan	Generate work plan from design	Planning phase
/build	Execute from existing task plan	Resume implementation
/review	Verify code against design docs	Post-implementation check

📚 Language-Agnostic Rules

Battle-tested best practices that work across all languages:

Coding Principles - Universal code quality standards
Testing Principles - TDD, coverage, test patterns
Architecture Patterns - Design decisions and trade-offs
Documentation Standards - Clear, maintainable docs

Problem: AI Coding Challenges

❌ Context exhaustion in long sessions
❌ Declining code quality over time
❌ Inconsistent patterns across team
❌ Manual quality checks and fixes

Solution: Specialized Workflows

✅ Fresh context per specialized agent
✅ Consistent quality through enforced rules
✅ Automated quality assurance
✅ Complete development lifecycle support

Sub Agents - Anthropic Documentation
Design Philosophy - 770K tokens without exhaustion

Intelligent Workflow Orchestration

Analysis: Assess task complexity and requirements
Planning: Generate appropriate documentation (PRD, Design Doc, Work Plan)
Execution: Specialized agents handle each phase autonomously
Quality: Automated testing, type checking, and error fixing
Review: Verify compliance and completeness
Commit: Clean, production-ready code

Time: 2 days → 30 TypeScript files with full coverage
Result: Production-deployed MCP server

Time: 1.5 days → Complete creative tool
Result: Multi-image blending, character consistency

Key Insight: Proper workflows + specialized agents = production-quality code at AI speed

/implement "Add user authentication with JWT" # → Automatically: # 1. Analyzes requirements # 2. Creates design documents # 3. Breaks down into tasks # 4. Implements with TDD # 5. Fixes all quality issues # 6. Reviews against design

/task "Fix validation error message" # → Direct implementation with quality checks

/review # → Verifies implementation matches design docs

claude-code-workflows/ ├── .claude-plugin/ # Plugin metadata │ └── plugin.json # Plugin configuration ├── agents/ # Specialized agent definitions │ ├── requirement-analyzer.md │ ├── technical-designer.md │ ├── quality-fixer.md │ ├── rules/ # Development rules │ │ ├── coding-principles.md │ │ ├── testing-principles.md │ │ ├── documentation-criteria.md │ │ └── architecture/ │ ├── guides/ # User documentation │ └── templates/ # Document templates ├── commands/ # Workflow commands │ ├── implement.md │ ├── task.md │ ├── design.md │ ├── plan.md │ ├── build.md │ └── review.md ├── LICENSE └── README.md

Q: Do I need to learn special commands? A: Just use /implement to start. The plugin handles the complexity automatically.

Q: What if there are errors? A: The quality-fixer agent automatically fixes most issues. If it can't, it provides clear guidance.

Q: Can I customize the workflows? A: Absolutely! After installation, edit rules in ~/.claude/agents/rules/ and agents in ~/.claude/agents/ to fit your team's needs.

MIT License - Free to use, modify, and distribute.

See LICENSE for full details.