Decodering: How I Use Claude Code — Agent Workflow

This talk walks through a real-world setup for using Claude Code as a multi-agent development system — not just a chat assistant, but a structured pipeline that handles requirements gathering, architectural design, implementation, code review, and documentation. I'll cover how custom agents hand off artifacts to each other, how persistent memory lets them learn across sessions, and how git worktrees enable parallel AI-assisted development without session conflicts. The focus is on practical patterns: what works, what's overkill, and how to build up this kind of workflow incrementally.

Target audience: Developers and tech leads who are already using Claude Code (or similar AI coding tools) and want to move beyond single-session prompting into repeatable, team-scale workflows. Familiarity with git and CLI tools assumed; no AI/ML background required.