Claude Code - Maximum Potential Configuration

New here? Type /start for quick reference or /health to check system status.

Cheatsheet (Copy-Paste Ready)

Start a Project

Use the idea-researcher agent to explore opportunities in [domain]
Use the project-planner agent to create a multi-terminal plan for [project]

Build Features

Use the elite-builder agent to implement [feature] with production quality
Use the fullstack-ml agent to build a complete AI application for [use case]

Quality Check

Use the code-reviewer agent to review my changes
Use the security-reviewer agent to audit the codebase
Use agents in parallel: code-reviewer + security-reviewer + typescript-reviewer

Quick Actions

/review          → Before any commit
/pr              → After feature complete
/debug           → When stuck
ultrathink       → For complex decisions

Quick Start Commands

Workflow Strategies

1. Explore → Plan → Code → Commit

1. Research: "Explore the codebase to understand [feature]"
2. Plan: "Use plan mode to design the implementation"
3. Think: "think hard about edge cases" or "ultrathink"
4. Code: Implement with verification
5. Review: "Use the code-reviewer agent"
6. Commit: Create meaningful commits

2. Test-Driven Development

1. Write failing test → commit
2. Implement minimal code → verify pass → commit
3. Refactor → verify pass → commit

3. Agent Delegation

"Use the architect agent to design the system"
"Use the security-reviewer agent to audit the code"
"Use agents in parallel: debugger + test-runner"

Multi-Agent Patterns

Parallel Review (Best for Quality)

Use agents in parallel:
- code-reviewer for quality
- security-reviewer for vulnerabilities
- typescript-reviewer for language patterns

Sequential Pipeline (Best for Features)

1. Use the planner agent to break down the task
2. Use the architect agent to design
3. Implement the code
4. Use the tdd-guide agent to verify tests
5. Use the code-reviewer agent for final review

Agent Teams (Pre-configured)

Located in ~/.claude/agent-teams/:

Usage: Use the product-quality-team to review this feature

Installed Components

Agents (85)

Core: architect, code-reviewer, debugger, refactorer, planner, tdd-guide

AI/ML Engineer Power Agents (NEW):

• idea-researcher - Deep research & idea generation
• project-planner - Multi-terminal parallel execution plans
• elite-builder - Best-in-class FAANG-quality implementation
• ml-architect - End-to-end ML system design
• fullstack-ml - Complete AI application builder
• innovation-tracker - Track new Claude features & AI news

Language Reviewers: typescript, python, go, rust, java, kotlin, cpp, flutter

Build Resolvers: build-error-resolver, typescript, python, go, rust, java, kotlin

AI/ML: ai-engineer, ml-engineer, data-scientist, nlp-engineer, computer-vision

Infrastructure: devops-engineer, cloud-architect, kubernetes, terraform, ansible, aws, azure, gcp

Security: security-reviewer, penetration-tester, security-architect

Database: database-reviewer, postgresql-expert, mongodb-expert

Skills (37)

api-design-reviewer, ci-cd-pipeline-builder, database-designer, docker-development, mcp-server-builder, agent-designer, autoresearch-agent, browser-automation, changelog-generator, dependency-auditor, focused-fix, git-worktree-manager

Rules (65)

Languages: TypeScript, Python, Go, Rust, Java, Kotlin, C++, C#, Swift, PHP, Perl

Categories: coding-style, patterns, security, testing, hooks

MCP Servers (9)

Hooks (5 - From Trail of Bits Best Practices)

Commands (12)

/start, /health, /pr, /tdd, /review, /ultrathink, /debug, /refactor, /quality-build, /add-language-rules, /feature-development, /database-migration

Security Features

Credential Protection (Trail of Bits)

Automatically blocked:

• ~/.ssh/* - SSH keys
• ~/.aws/credentials - AWS secrets
• ~/.kube/config - Kubernetes config
• ~/.gnupg/* - GPG keys
• **/.env - Environment files
• **/secrets.json - Secret files

Dangerous Command Blocking

PreToolUse hooks block:

• rm -rf → Suggests using trash
• git push origin main → Requires feature branch
• git push --force → Suggests --force-with-lease
• .env writes → Use environment variables

Audit Logging

All Bash, Edit, Write operations logged to: ~/.claude/logs/audit-YYYY-MM-DD.jsonl

Model Selection

Commands: /model, /fast (2.5x speed, 6x cost)

Thinking Modes

• think - Basic extended thinking
• think hard - Deeper analysis
• ultrathink - Maximum reasoning depth
• Plan mode: Shift+Tab to toggle

Best Practices

Code Quality

• Simple > clever
• Follow existing patterns
• Test after changes
• Handle errors gracefully

Security

• No hardcoded secrets
• Validate all inputs
• Parameterized queries
• OWASP guidelines

Performance

• Research first for complex tasks
• Use subagents for verification
• Iterate 2-3 rounds for quality
• /compact at 50% context usage

Key Shortcuts

Agent Team Mode (Experimental)

Enable with: CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1

Allows multi-agent collaboration with lead agent coordinating specialists.

Documentation

Sources & Inspiration

• Trail of Bits claude-code-config
• disler/claude-code-hooks-mastery
• rohitg00/awesome-claude-code-toolkit
• affaan-m/everything-claude-code

Last validated: 2026-03-29 | Status: Production Ready