THE AGENT SPECTRUM

From readiness
to agent systems.

Check your prerequisites, learn the basics, then build reliable agents and coordinated systems.

01
BEFORE YOU JOIN

Prerequisites

Knowledge you must already have before joining. These are entry requirements, not lessons we teach.

Entry requirementsBefore Basics

Outcomes

  • Explain a task and its inputs
  • Define a correct result
  • Make a small webpage change

Topics

HTML, CSS & JavaScript

Recognize structure, styling, and behavior; make a simple edit.

Express an idea clearly

Describe the user, problem, desired result, and constraints.

Web versus mobile

Understand browser-based apps and installed mobile apps.

Break a task into steps

Identify inputs, outputs, and examples of a successful result.

Programming literacy

Understand variables, conditions, loops, files, and folders at a basic level.

Check the result

Compare an answer with evidence and explain what is missing.

02
LEARN THE FOUNDATIONS

Basics

Understand the tools and systems behind agents. Build and deploy a small app that uses a model or a tool.

Taught stage 1Understand → deploy

Outcomes

  • Publish a working app
  • Connect a model or tool
  • Debug and test outcomes

Topics

Prompts & context

Write clear prompts; supply relevant context and understand model limitations.

Local & remote development

Use a terminal, run locally, and understand remote sessions and environments.

Git & GitHub

Track changes, use branches, review a diff, and collaborate through pull requests.

Frontend, backend & data

Trace a request from the interface through an API to a database.

Models, tools & agents

Distinguish a model response, a fixed workflow, and an agent choosing its next action.

Skills & hooks

Package reusable skills and trigger actions at defined lifecycle events.

APIs & structured data

Connect tools, read JSON, and handle credentials and permissions.

Deploy, debug & evaluate

Publish an app, read logs, test expected outputs, and measure usage costs.

03
BUILD AGENT SYSTEMS

Advanced

Build one reliable agent, then loops, a complete service, and coordinated agents. Improve the system through measured experiments.

Taught stage 2One agent → a system

Outcomes

  • Deploy a complete agent service
  • Coordinate agents and recover failures
  • Demonstrate measured improvement

Topics

A single agent

Give an agent a goal, tools, memory, permissions, and an observable result.

Agent loops

Plan, act, evaluate, and repeat with explicit stop conditions and budgets.

A complete service

Connect interface, backend, storage, authentication, deployment, and monitoring.

Coordinated agents & swarms

Delegate work, exchange context, run tasks in parallel, and resolve conflicting results.

Reliability & evaluation

Use test sets, traces, retries, queues, and recovery; compare quality and cost with a single agent.

Measured self-improvement

Propose prompt, skill, or code changes; test on held-out cases, then keep or roll back.

Recursive improvement: experimental

Explore improving the improvement process itself. Require independent evaluation; progress is not guaranteed.

Part of 10claws