AGENT MENTOR · OPEN COURSES
Agent changes, turned into courses you can use right away.
Only topics that are directly about Agents and still moving fast. Courses are free; exercises, code, and selected text copy straight into Claude Code, Codex, or your Agent of choice.
12 courses · 72 lessons · ~25 hours · read in order
Start using agents
Get an agent to finish something real on its own
01Claude Code Skills: Build Your Own AI WorkflowsStart hereTurn your repetitive workflows into Skills that Claude Code follows on its own.02Designing Agent Workflows: From One-Off Conversations to Multi-Step AutomationBreak a multi-step job into an orchestrated pipeline that survives interruptions and picks up where it left off.03Prompt Engineering Basics: How to Write Effective InstructionsWrite prompts that state the requirement clearly the first time, and debug them systematically.04Agent Tool Calling: Getting Agents to Actually Do ThingsRead the tool-calling protocol and design tool interfaces the model picks right, fills right, and can't overreach with.
Make the workflow hold
Reuse it, budget its context, verify what comes back
05Agent Memory and StateTell memory from state, and give an agent memory that survives across sessions without being poisoned.06Multi-Agent CollaborationSplit work across several agents without chaos: division of labor, delegation prompts, producer-reviewer.07Agent Harness Fundamentals: Loops and ControlHand-write a stop_reason-driven agent loop and fit it with four stop valves.08Context Engineering: Spending Finite Attention Where It CountsDecide what goes into — and out of — the context every turn, so long tasks don't blow the window.09State Management and Persistence: Making Long Tasks Survive InterruptionMake long tasks survive interruption: checkpoints, an effects ledger, resume from the break.10Verification and Quality Assurance: Don't Let 'Looks Right' Slip ThroughBuild an eval track for your agent's output that separates "looks done" from "is done.
Build it into the system
Repositories, real tools, multiple agents
11Observability and Debugging: Seeing Every Step Your Agent TakesFit your agent with logs, metrics, and a trace tree, so a failure traces back to the exact message.12From Loops to Graphs: Orchestration Engineering for Agent SystemsUpgrade your single-loop harness into an orchestration script with routing, fan-out, and a review loop.
Courses, exercises, and sources live locally — ready for your own Agent. About Agent Mentor ↗