AOL 5 - Hooks | Resource Hub | Agent Builder Academy

The short version

At AOL 5, intent becomes automated. Hooks are triggers that run automatically at specific moments - before a task starts, after a file changes, when a build completes. They connect the skills you built at AOL 4 to the events that should activate them. The result is a setup that self-corrects without you being the one to notice something went wrong. In Claude Code these are called Hooks. Other tools implement the same concept differently - the principle is the same.

How hooks work

A hook watches for an event. When the event happens, it runs a command or a skill. You do not have to remember to check - the system checks for you. Here are some examples.

Before every session starts

Run the design check skill. Flag anything that does not match the system before new work begins.

After a component is created

Verify the component follows the naming convention and is added to the barrel export.

Before a commit is made

Run the build and type check. Block the commit if errors exist.

After a file is modified

Check whether the change is within the approved scope. Alert if unexpected files were touched.

What this changes

At AOL 4, you invoke skills manually. At AOL 5, the skills are invoked by events. The difference is significant. You no longer rely on remembering to check - the project checks itself. The agent becomes proactive instead of reactive.

AOL 4

You run the check when you remember to.

AOL 5

The system runs the check every time the relevant event happens.

AOL 5 is not automation for its own sake

Hooks amplify what is already there. Solid skills become reliable automation. Inconsistent skills become automated inconsistency. Build the manual version first. Add the hook when you trust the result.

Your role at AOL 5

Before

Directing every action. Invoking checks manually. Noticing when something goes wrong.

After

Defining what runs automatically. Reviewing what the system flags. Spending time on decisions, not checks.

The full construction site

Each AOL layer adds one thing. Together they form a project where the agent knows where it is, knows how things should work, can run your processes on command, and checks its own work automatically. You build the layers in order. You do not skip to hooks.

AOL 2 CLAUDE.md The Map - tells the agent where it is

AOL 3 /docs The Blueprint - tells the agent how it should look

AOL 4 Skills The Tools - give the agent the right tool for each job

AOL 5 Hooks The Smart Site - the site checks itself

Starter file

Hooks Setup Guide Common hook patterns with working JSON config examples - auto-format, type check, and test triggers.

The AI Setup Snapshot shows your current AOL alongside your APL and AAL. It tells you exactly which layer to build next.

Take the AI Setup Snapshot