Learn Claude Code

Lộ trình học

s01 đến s17: thiết kế Agent Harness tuần tự

Chú giải tầng

Tools & Execution
Planning & Control
Memory Management
Concurrency & Scheduling
Multi-Agent Platform
01
s01Minimal model/tool loop

The Agent LoopOne Loop Is All You Need

106 dòng mã1 công cụ

“The smallest useful agent is a loop that calls the model, runs tools, and feeds results back.”

Tìm hiểu thêm
02
s02Tool dispatch map

Tool UseAdd a Tool, Add Just One Line

149 dòng mã5 công cụ

“The loop stays stable while capabilities register into a dispatch table.”

Tìm hiểu thêm
03
s03Permission gate

PermissionCheck Permissions Before Execution

185 dòng mã5 công cụ

“Dangerous actions need a harness decision point before the shell runs.”

Tìm hiểu thêm
04
s04Lifecycle hooks

HooksHang on the Loop, Don't Write into It

207 dòng mã5 công cụ

“Cross-cutting behavior belongs around the loop, not tangled inside it.”

Tìm hiểu thêm
05
s05Todo manager

TodoWriteAn Agent Without a Plan Drifts Off Course

284 dòng mã6 công cụ

“Explicit plans keep long-running work visible and correctable.”

Tìm hiểu thêm
06
s06Isolated subtask context

SubagentBreak Large Tasks into Small Ones with Clean Context

291 dòng mã6 công cụ

“Subagents give each subtask a clean message history while preserving the main thread.”

Tìm hiểu thêm
07
s07On-demand skill loader

Skill LoadingLoad Only When Needed

306 dòng mã6 công cụ

“Inject specialized knowledge only when the task actually needs it.”

Tìm hiểu thêm
08
s08Context compaction

Context CompactContext Will Fill Up

503 dòng mã5 công cụ

“Compression keeps the conversation usable when the context window gets crowded.”

Tìm hiểu thêm
09
s09Durable memory layer

MemoryKeep a Layer That Doesn't Lose Details

679 dòng mã5 công cụ

“Some facts should survive summarization and future sessions.”

Tìm hiểu thêm
10
s10Task board

Task SystemBreak Big Goals into Small Tasks

466 dòng mã11 công cụ

“A task graph turns vague goals into ordered, observable work.”

Tìm hiểu thêm
11
s11Background execution

Background TasksSlow Operations Go to the Background

404 dòng mã5 công cụ

“The agent can keep reasoning while slow work completes elsewhere.”

Tìm hiểu thêm
12
s12Scheduled task creation

Cron SchedulerProducing Work on a Schedule

642 dòng mã5 công cụ

“Recurring work should be created by the harness, not remembered by the model.”

Tìm hiểu thêm
13
s13Team runtime with task-bound worktrees

Agent Team RuntimePersistent Teammates, Atomic Claims, Task-Bound Worktrees

1592 dòng mã5 công cụ

“Persistent teammates can reliably discover and execute parallel work when the runtime owns messaging, atomic claims, and task-bound working directories.”

Tìm hiểu thêm
14
s14MCP tool bridge

MCP ToolsExternal Tools, Standard Protocol

444 dòng mã5 công cụ

“External services can become agent tools through a standard discovery and call protocol.”

Tìm hiểu thêm
15
s15Integrated harness

Integrated HarnessMany Mechanisms, One Loop

2770 dòng mã26 công cụ

“The integrated harness is still one loop, surrounded by the systems introduced across the course.”

Tìm hiểu thêm
16
s16Resumable workflow runtime

Workflow RuntimeScripts Own Fixed Orchestration

725 dòng mã27 công cụ

“When orchestration has a fixed shape, code can make it parallel, deterministic, and resumable.”

Tìm hiểu thêm
17
s17Goal completion gate

Goal LoopIndependent Evaluation Decides When to Stop

794 dòng mã5 công cụ

“A durable goal keeps the loop working until an independent evaluator finds the completion condition satisfied in the conversation.”

Tìm hiểu thêm

Tăng số dòng mã

s01
106
s02
149
s03
185
s04
207
s05
284
s06
291
s07
306
s08
503
s09
679
s10
466
s11
404
s12
642
s13
1592
s14
444
s15
2770
s16
725
s17
794