Most people use 10% of Claude Code’s capability. This course fixes that. You’ll master the full stack — hooks, CLAUDE.md architecture, custom subagents, multi-agent workflows, the Agent SDK, and production hardening — all in your real codebase, not toy examples. Every session builds toward a capstone architecture document and a production-hardened agent you can actually deploy.
Levels: Remember · Understand · Apply · Analyze · Evaluate · Create — highest demands most original thinking.
Build the mental model: prompting (ask→answer) vs agentic (goal→plan→tools→verify). Map the three paradigms — interactive, autonomous, hybrid.
Exercise: Your Agent BaselineGap-filling for daily CC users: /effort levels, /voice, /loop automation, multi-instance patterns, Agent Teams, and breaking changes.
Exercise: CC Power User PlaybookConfiguration deep dive: CLAUDE.md cascade, settings.json levels, instruction budgets, skills budgets, auto-memory, and @import syntax.
Exercise: CLAUDE.md RefactorExtension types: plugins, skills, hooks, agents, MCP servers, desktop extensions. Evaluate the ecosystem and assess supply chain security risks.
Exercise: Plugin Ecosystem AuditConfigure custom subagents with tool restrictions and model routing. Build a 3-agent code review system.
Exercise: 3-Agent Code Review System22 hook events with decision control types. HTTP hooks, security context (real CVEs), and pipeline architecture patterns.
Exercise: Production Hooks + Pipeline DesignAgent SDK deep dive: agent loop, tool execution, compaction API, session persistence. Build agents you can deploy outside Claude Code.
Exercise: Two Standalone Agents/loop cost model and compaction between iterations. Evaluate Ralph Loop, Claude-Mem Endless Mode, and when autonomous loops are worth it.
Exercise: Autonomous Loop Trial RunAgent Teams, WebMCP W3C Draft, cross-device orchestration with Dispatch and Codex Cloud. Cost reality: 3–5x tokens.
Exercise: Multi-Agent Course BuilderHarden agents for production: MCP security posture, cost management with /effort and prompt caching, testing, and monitoring.
Exercise: AI Uni Agent Architecture + Production AgentUnderstanding agent loops, the three paradigms (interactive, autonomous, hybrid), human-in-the-loop models, and how Claude Code fits in the agentic landscape.
CLAUDE.md hierarchy, settings.json load order, instruction budgets, skills budgets, @import syntax, auto-memory, and the WHY/WHAT/HOW pattern.
Evaluating plugins and skills, reading SKILL.md, cross-platform compatibility, supply chain security assessment, and building custom skills.
Custom subagents with YAML frontmatter, model routing strategies, Agent Teams with shared task boards, Dispatch orchestration, and the WebMCP standard.
All 22 hook events, HTTP hooks, pipeline architecture (sequential, DAG, event-driven, swarm), autonomous loops, CI/CD integration, and cost tracking.
MCP threat landscape (30+ CVEs), permission modes, Claude Code Security scanning, cost management via /effort and prompt caching, Enterprise Analytics API monitoring.
AI Uni Agent Architecture — A comprehensive architecture document covering every repeatable process mapped to agents, skills, and automations. Includes .claude folder architecture, MCP servers, hooks, cost projections with /effort routing, security posture informed by real CVE data, and cross-tool strategy. Plus one production-hardened agent with cost controls, security audit, verification hooks, error recovery, and monitoring. Assembled across all sessions, finalized in Session 10.
Anthropic’s agentic coding tool. The primary environment for every session — hooks, subagents, skills, and configuration all happen here.
TypeScript SDK for building standalone agents with tool execution, session persistence, compaction handling, and model routing.
Model Context Protocol for tool connectivity. Build MCP servers, evaluate security, and understand the protocol that connects the ecosystem.
Anthropic’s desktop orchestration layer for cross-device agent workflows, background task monitoring, and multi-agent coordination.
Enterprise/Team vulnerability scanning feature. Deep reasoning analysis that traces data flows, maps dependencies, and simulates attack paths.
Superpowers, Claude-Mem, Ralph Loop, Ruflo — ecosystem plugins you’ll evaluate, compare, and integrate into your workflows.
Most people use 10% of Claude Code. Session 1 shows you the other 90%.
Join the waitlist