Multi-agent architecture in Claude Code that enables parallel task execution through specialized AI workers with isolated context windows, overcoming single-agent limitations for complex operations.

Explores user intent, requirements and design before implementation through collaborative dialogue.

Dispatch multiple independent agents to work on unrelated tasks concurrently, maximizing efficiency.

Load implementation plans, execute tasks in batches with review checkpoints between batches.

Guide completion of development work with structured options for merge, PR, or cleanup after tests pass.

Technical evaluation of code review feedback - verify before implementing, ask before assuming.

Dispatch code-reviewer subagent to catch issues before they cascade - review early, review often.

Execute plans by dispatching fresh subagent per task with two-stage review - spec compliance then code quality.

Always find root cause before attempting fixes - systematic approach to debugging any technical issue.

Write tests first, watch them fail, then write minimal code to pass - ensuring code quality through TDD methodology.

Creates isolated git workspaces for feature work with smart directory selection and safety verification.

Meta-skill that establishes how to find and use Skills - invoke relevant Skills before ANY response.

Evidence before claims - always run verification commands and confirm output before making completion claims.

Create comprehensive implementation plans with bite-sized tasks assuming zero context, before touching code.

Test-Driven Development applied to process documentation - create and verify Skills work before deployment.