Reflection Day

Today is a rest & review day. Look back at your P1-P2 journey:
Which SDK concepts felt intuitive? Where did you get stuck?
Write a 3-sentence summary of what you've learned so far.

Write Reflection Review Flashcards Skip Today

Track 1 · Awakening

P2 · Tool Use

~3 hours · 3 lessons + 3 exercises · +500 XP

Function Calling Mental Model

Tool Design Principles

Atomicity, clear descriptions, helpful errors

Lesson · +50 XP

Continue

Tool Orchestration

When to call, when not to, result into context

Exercise: Implement Calculator Tool

Exercise: File & Search Tool Mocks

Exercise: Build agentLoop(msg, tools)

All Tracks

Track 1 · Awakening P1 Agent Basics · P2 Tool Use · P3 Prompts

1/3

Gate: L2 Apprentice

Complete P3 + run query() with outputFormat

Requirements:
• Complete all 3 lessons in Track 1
• Pass the P3 structured output exercise
• Successfully call query() with Zod schema validation

Continue to P3

Track 2 · Forging P4 Memory · P5 Eval & Routing · P6 Patterns & Capstone

0/3

Gate: L3 Builder

Pass P6 Capstone — build a mini agent framework using 7+ SDK features

P6 Capstone Project:
Build a mini agent framework demonstrating mastery of:
• query() + resume session management
• allowedTools + canUseTool permission control
• outputFormat structured output with Zod
• hooks.PreToolUse + PostToolUse lifecycle
• Error handling + graceful degradation

View Gallery Start P6

Track 3 · Expedition MCP · RAG · Multi-Agent · Code Gen · HITL · Production

Pick 2+ Choose your path

Recommended Routes

Full-Stack Agent

P7 + P10 + P12

MCP, Code Gen, Production deployment

SDK Skills: createSdkMcpServer() enableFileCheckpointing rewindFiles() sandbox{}
Outcome: Build a code generation agent with MCP tool ecosystem, file safety net, and production-ready sandbox.
~14 hours · Best for developers building coding assistants.

AI Engineer

P8 + P9 + P12

RAG, Multi-Agent, Production systems

SDK Skills: agents{} AgentDefinition maxTurns hooks.SubagentStart/Stop
Outcome: Build multi-agent RAG pipelines with orchestration, relevance scoring, and production monitoring.
~16 hours · Best for AI/ML engineers building knowledge systems.

Safety-First

P7 + P11 + P12

MCP, HITL guardrails, Production safety

SDK Skills: permissionMode canUseTool AskUserQuestion hooks.PermissionRequest
Outcome: Build permission-gated agents with audit trails, sandbox enforcement, and human oversight.
~12 hours · Best for engineers focused on trust & safety.

Quick Access

Flashcards 5 due Exercises 6 Chat Sim

Today's XP 105 remaining

2 lessons +100

Flashcards +30

Daily quest +20

Focus bonus +45

Total 195 / 300