CWP Claude Commands

A collection of custom slash commands for Claude Code to streamline software development workflows.

Available Commands

/cwp:prime

Purpose: Initialize project context by reading key project files.

Usage: /cwp:prime

What it does:

• Runs git ls-files to list all tracked files in the repository
• Reads essential project files:
  • README.md - Project overview and documentation
  • CLAUDE.md - Claude-specific guidelines and patterns
  • package.json - Project dependencies and scripts
  • ai_docs/Rules.md - AI-specific rules and conventions

Use this command at the start of a session to give Claude full context about your project structure and conventions.

---

/cwp:commit [all]

Purpose: Create a well-formatted git commit with an AI-generated commit message.

Usage:

• /cwp:commit - Commits currently staged changes
• /cwp:commit all - Stages and commits all changed files

Model: Uses Haiku for fast execution

What it does:

• Analyzes current changes in the repository
• Generates a descriptive commit message
• Creates a commit with the generated message
• If all argument is passed, includes ALL changed files

---

/cwp:implement [path|prompt] [--parallel]

Purpose: Execute complex feature implementation with optional parallel development using multiple coder agents.

Usage:

• /cwp:implement "Add user authentication" - Sequential implementation
• /cwp:implement path/to/tasks.md --parallel - Parallel implementation with multiple agents

What it does:

• Analyzes the implementation requirements
• Breaks down work into tasks
• Spawns coder subagent(s) to implement features
• If --parallel flag is provided: creates 2-5 parallel coder agents for independent tasks
• Each coder must update task checkboxes and pass bun check (or npm/yarn)
• Runs quality gates and verification after completion

Quality Gates:

• All task checkboxes must be marked complete
• Build and type checking must pass with zero errors
• All tests must pass
• No any types in TypeScript code
• Code must follow project patterns

---

/cwp:tech-stack [<path-to-project-brief-or-PRD>]

Purpose: Analyze requirements and recommend/document the technology stack.

Usage:

• /cwp:tech-stack - Analyzes current codebase and creates tech stack docs
• /cwp:tech-stack path/to/requirements.md - Analyzes requirements first

What it does:

1. Analyzes Requirements (if path provided):
   • Reads project brief or PRD
   • Identifies scalability, integration, and performance requirements
2. Analyzes Existing Stack:
   • Maps current technologies
   • Identifies reusable components
   • Flags breaking changes needed
3. Recommends Stack:
   • Frontend framework & libraries
   • Backend framework & runtime
   • Database & cache solutions
   • Infrastructure & deployment
   • Development tools & testing
4. Documents Decisions:
   • Rationale for each choice
   • Trade-offs vs alternatives
   • Compatibility matrix
   • Migration path (if updating)

Output: Creates tech-stack.md with complete technology stack documentation

---

/cwp:review-code

Purpose: Perform a comprehensive code review of the latest commit.

Usage: /cwp:review-code

What it does:

1. Identifies the Commit: Retrieves and displays the most recent commit hash and message
2. Conducts Code Review:
   • Analyzes changes in the commit
   • Provides three specific, actionable suggestions for improvement
   • Focuses on performance, readability, maintainability, and security
3. Considers Edge Cases:
   • Identifies missing or unhandled edge cases
   • Recommends updates to handle these scenarios

Provides thorough, professional feedback as if submitting a formal pull request review.

---

/cwp:prompt-optimizer <prompt>

Purpose: Optimize prompts for better AI model performance and efficiency.

Usage: /cwp:prompt-optimizer "Your prompt here"

Model: Uses Sonnet for advanced reasoning

What it does:

1. Prompt Engineering:
   • Applies chain-of-thought reasoning
   • Adds few-shot examples
   • Implements role-based instructions
   • Uses clear delimiters and formatting
   • Adds output format specifications
2. Context Optimization:
   • Minimizes token usage
   • Structures information hierarchically
   • Removes redundant information
   • Adds relevant context
   • Uses compression techniques
3. Performance Testing:
   • Creates prompt variants
   • Designs evaluation criteria
   • Tests edge cases
   • Measures consistency
4. Model-Specific Optimization:
   • GPT-4 best practices
   • Claude optimization techniques
   • Prompt chaining strategies
   • Temperature/parameter tuning
   • Token budget management
5. RAG Integration:
   • Context window management
   • Retrieval query optimization
   • Chunk size recommendations
   • Embedding strategies
6. Production Considerations:
   • Prompt versioning
   • A/B testing framework
   • Monitoring metrics
   • Fallback strategies
   • Cost optimization

Output: Optimized prompts with explanations, evaluation metrics, and testing strategies.

---

/cwp:error-analisys <error-context>

Purpose: Comprehensive error analysis and resolution guidance.

Usage: /cwp:error-analisys "error description or path"

Model: Uses Sonnet for deep analysis

What it does:

1. Error Pattern Analysis:
   • Categorizes error types
   • Identifies root causes
   • Traces error propagation
   • Analyzes error frequency
   • Correlates with system events
2. Debugging Strategy:
   • Stack trace analysis
   • Variable state inspection
   • Execution flow tracing
   • Memory dump analysis
   • Race condition detection
3. Error Handling Improvements:
   • Custom exception classes
   • Error boundary implementation
   • Retry logic with backoff
   • Circuit breaker patterns
   • Graceful degradation
4. Logging Enhancement:
   • Structured logging setup
   • Correlation ID implementation
   • Log aggregation strategy
   • Debug vs production logging
   • Sensitive data masking
5. Monitoring Integration:
   • Sentry/Rollbar setup
   • Error alerting rules
   • Error dashboards
   • Trend analysis
6. Recovery Mechanisms:
   • Automatic recovery procedures
   • Data consistency checks
   • Rollback strategies
7. Prevention Strategies:
   • Input validation
   • Type safety improvements
   • Contract testing
   • Defensive programming

Output: Specific fixes, preventive measures, test cases, and long-term reliability improvements.

---

Installation

These commands are located in .claude/commands/cwp/ and are automatically available in Claude Code when you use the cwp: namespace.

Project Structure

.claude/commands/cwp/

Custom slash commands for Claude Code. Each command is a .md file with frontmatter and a prompt.

.claude/agents/

Custom agent definitions that extend Claude Code's capabilities:

• gemini-cli.md - A routing subagent that integrates Google's Gemini model for long-context reasoning and deep code analysis. It acts as a pure pass-through router, collecting file paths and passing tasks to Gemini via the gemini-cli wrapper script.

scripts/

Utility scripts for enhanced functionality:

• gemini-cli/ - TypeScript wrapper for calling Google's Gemini model from Claude subagents
  • gemini-cli.ts - Main CLI wrapper that formats requests for Gemini and normalizes responses
  • build.sh - Build script to compile and install the CLI tool to ~/.local/bin
  • Usage: echo '<payload-json>' | gemini-cli --task "<task>" --stdin
  • Requirements: Node.js, ts-node (or bun), and the official gemini CLI installed and authenticated

Contributing

To add new commands:

1. Create a new .md file in .claude/commands/cwp/
2. Add frontmatter with description, argument-hint, and optional model specification
3. Write the command prompt
4. Update this README with documentation

To add new agents:

1. Create a new .md file in .claude/agents/
2. Add frontmatter with name, description, tools, model, and permissionMode
3. Define the agent's behavior and constraints
4. Update this README with documentation

Command Structure

Each command file follows this structure:

---
description: Brief description of the command
argument-hint: [optional-arg] [--flags]
model: sonnet|haiku|opus  # Optional, defaults to sonnet
---

# Command prompt content here

Agent Structure

Each agent file follows this structure:

---
name: agent-name
description: >
  Brief description of what the agent does
tools: Read, Write, Bash
model: sonnet|haiku|opus
permissionMode: default|strict
---

# Agent instructions and behavior definition here