ORI (Optimize-Research-Implement) for Claude Code

A powerful slash command for Claude Code that executes autonomous multi-phase workflows for complex software engineering tasks. Based on the ORI framework with intelligent model selection, quality gates, and comprehensive error handling.

Features

• Multi-Phase Workflow - 5 phases from strategy to documentation
• Intelligent Model Selection - Automatic selection of Opus/Sonnet/Haiku based on complexity
• Deep Research - Web search, documentation analysis, codebase examination
• Security Verification - OWASP Top 10 checks, vulnerability assessment
• Quality Gates - Validation between each phase ensures high-quality output
• Error Recovery - Automatic rollback and alternative approaches
• Context Preservation - Handoff packets maintain full context between phases
• Zero Setup - Works immediately as a slash command

What is ORI?

ORI is a structured workflow framework that breaks down complex tasks into manageable phases:

Phase 0: STRATEGY    → Analyze task, design research plan
Phase 1: RESEARCH    → Deep dive into docs, examples, best practices
Phase 2: VERIFY      → Cross-validate, security review, feasibility check
Phase 3: IMPLEMENT   → Execute solution with tests and error handling
Phase 4: DOCUMENT    → Update README, CHANGELOG, create summary

Each phase uses the optimal AI model:

• Opus for complex reasoning and strategic planning
• Sonnet for balanced implementation and verification
• Haiku for fast documentation updates

Installation

Quick Install (Recommended)

1. Download the command file:

curl -o ~/.claude/commands/ori.md \
  https://raw.githubusercontent.com/chrisabra-co/ori-command/main/.claude/commands/ori.md

2. Restart Claude Code

3. The /ori command is now available globally!

Manual Install

1. Clone the repository:

git clone https://github.com/chrisabra-co/ori-command.git
cd ori-command

2. Copy the command to your Claude Code configuration:

mkdir -p ~/.claude/commands
cp .claude/commands/ori.md ~/.claude/commands/

3. Restart Claude Code

Usage

Basic Usage

# Execute any complex task
/ori [your task description]

# Examples
/ori implement JWT authentication for the Express API
/ori add rate limiting to all API endpoints
/ori refactor the user service to use dependency injection
/ori optimize database queries in the analytics module

Example Workflow

Input:

/ori implement JWT authentication for Express API with refresh tokens

Phase 0: Strategy (Opus)

• Analyzes task complexity: HIGH
• Domain: Backend/Security
• Risk flags: security, authentication
• Recommends: Opus for research, Sonnet for implementation
• Designs research strategy with 5 key questions

Phase 1: Research (Opus - complex task)

• Searches official JWT documentation
• Reviews Express authentication patterns
• Finds refresh token implementation examples
• Examines security best practices (OWASP)
• Identifies required packages: jsonwebtoken, bcrypt

Phase 2: Verification (Sonnet)

• Cross-validates JWT approach is current
• Security review:
  • Token storage (httpOnly cookies)
  • Secret management (environment variables)
  • Token expiration (access: 15min, refresh: 7days)
  • Refresh token rotation
• Performance: Minimal impact
• Decision: APPROVED with security enhancements

Phase 3: Implementation (Sonnet)

• Creates auth middleware
• Implements login/refresh endpoints
• Adds token generation/validation
• Writes comprehensive tests
• Includes error handling and logging

Phase 4: Documentation (Haiku)

• Updates README with auth setup
• Adds CHANGELOG entry
• Creates usage examples
• Documents environment variables

Result: Complete JWT authentication system with refresh tokens, security best practices, tests, and documentation.

How It Works

Phase 0: Research Strategy (Opus)

Analyzes your request to understand:

• Complexity: LOW/MEDIUM/HIGH
• Domain: Code, data, infrastructure, UX, etc.
• Novelty: Established pattern vs. new problem
• Risk Flags: Security, privacy, compliance concerns

Designs optimal research strategy:

• Primary questions to answer
• Information sources to consult
• Search queries to execute
• Validation criteria

Selects models for each phase based on complexity.

Phase 1: Deep Research (Dynamic)

Gathers comprehensive information:

• Web Search: Latest documentation, best practices
• Code Examples: GitHub patterns, working implementations
• Codebase Analysis: Existing patterns, related components
• Synthesis: Best approach, alternatives, assumptions

Uses Opus for complex/novel tasks, Sonnet for standard tasks.

Phase 2: Verification (Sonnet)

Validates research findings:

• Cross-Validation: Confirm approach is current
• Security Review: OWASP Top 10, vulnerabilities
• Performance: Scalability, bottlenecks
• Feasibility: Dependencies, compatibility, complexity

Makes final decision on approach.

Phase 3: Implementation (Sonnet/Haiku)

Executes the solution:

• Safety First: Git check, backup, branch creation
• Implementation: Step-by-step with validation
• Testing: Unit tests, integration tests, edge cases
• Error Handling: Automatic recovery or rollback

Uses Sonnet for complex code, Haiku for simple edits.

Phase 4: Documentation (Haiku)

Updates all documentation:

• README: Feature docs, setup, usage examples
• CHANGELOG: Version entry, changes, breaking changes
• Summary: Implementation report, next steps
• Comments: Inline documentation, TODOs

Quality Gates

Each phase must pass validation before proceeding:

Phase 0 → 1:

• Complete strategy
• Risk flags identified
• Model selection justified

Phase 1 → 2:

• Minimum 2 sources consulted
• Primary approach identified
• Assumptions documented

Phase 2 → 3:

• Security review passed
• No critical blockers
• Clear implementation plan

Phase 3 → 4:

• Core implementation complete
• Critical tests passing
• No unresolved errors

Model Selection Logic

When to Use Opus

• Strategic planning and research design
• Complex multi-step reasoning
• Novel or ambiguous problems
• High-stakes architectural decisions

When to Use Sonnet

• Most implementation tasks
• Code generation and refactoring
• Validation and verification
• Balanced performance/cost

When to Use Haiku

• Simple file edits
• Documentation updates
• Formatting and style fixes
• Quick, straightforward tasks

Benefits

• Comprehensive Solutions - Nothing overlooked, all aspects covered
• Security First - Automatic security review on all implementations
• Best Practices - Research ensures current, recommended approaches
• Cost Optimized - Intelligent model selection minimizes cost
• Quality Assured - Multi-phase validation and quality gates
• Error Resilient - Automatic recovery and rollback strategies
• Well Documented - Complete documentation updates included
• Learning Tool - See expert workflow for complex tasks

When to Use ORI

Best For:

• Complex features requiring research
• Security-sensitive implementations
• New technology/framework adoption
• Architectural changes
• Performance optimization
• Multi-component integrations

Not Needed For:

• Simple bug fixes
• Trivial edits
• Well-understood tasks
• Quick experiments

Example Use Cases

Authentication Implementation

/ori implement OAuth2 authentication with Google provider

Performance Optimization

/ori optimize the dashboard query that's taking 5+ seconds

Feature Addition

/ori add real-time notifications using WebSockets

Security Hardening

/ori add rate limiting and request validation to all API endpoints

Architecture Refactoring

/ori refactor the monolith to microservices starting with user service

Configuration

The ORI command is self-contained and requires no configuration. However, you can customize behavior by:

1. Model Overrides - Edit .claude/commands/ori.md to change model selections
2. Phase Customization - Adjust research depth, verification criteria
3. Quality Gates - Modify validation requirements between phases

Comparison with MCP Server

Slash Command Version (This Project)

• ✅ Zero setup, works immediately
• ✅ No separate server process
• ✅ Lightweight and fast
• ✅ Perfect for Claude Code users
• ❌ No persistent logging
• ❌ No programmatic API access

MCP Server Version

• ✅ Persistent execution logs
• ✅ Programmatic tool access
• ✅ SME quality gate agents
• ✅ Configurable workflows
• ❌ Requires MCP setup
• ❌ Additional complexity

Troubleshooting

Command Not Found

• Ensure file is at ~/.claude/commands/ori.md
• Restart Claude Code completely
• Check file permissions (should be readable)

Phase Failures

• Review error messages in output
• Check if quality gates were met
• Verify required tools available (git, etc.)

Model Selection Issues

• Phase 0 may select different model than expected
• This is intentional based on complexity analysis
• Trust the strategic analysis

Contributing

Contributions welcome! Areas for improvement:

• Additional domain-specific research strategies
• More sophisticated model selection criteria
• Enhanced security verification rules
• Extended quality gate validations
• Custom phase templates

License

MIT License - see LICENSE file for details

Credits

Based on the MCP Server ORI by Grandin Harrison. This slash command version was adapted for Claude Code to provide a lightweight, zero-setup implementation of the ORI workflow framework.

Special thanks to Grandin Harrison for the original ORI framework and workflow design.

Support

• Issues: GitHub Issues
• Original Framework: MCP Server ORI
• Documentation: See the command file for complete workflow details

---

Made for complex software engineering tasks with Claude Code