Do you get your agent to brighten up your day?

You open a new AI agent session dozens of times a day. Every time, its first reply starts with the same polite, forgettable line. It's fine, but it's also the most repetitive part of your working day.

You also have no quick way to know whether the agent loaded your global instructions. You find out an hour later, when it ignores a rule you thought it knew.

One line in your global instructions file fixes both. Ask the agent to start its first reply in every session by greeting you with a funny nickname. You get a small laugh before the real work starts, and the greeting shows that the agent read your instructions.

Add one line to your global instructions

Put the instruction in your personal global file, not the project's AGENTS.md. Your teammates didn't sign up to be called "Captain Merge Conflict". See Do you keep universal rules for your AI agent in a global instructions file?

❌ Figure: Bad example - The same greeting every agent gives, every session

When starting a new interactive session, greet Daniel Mackay with a
funny nickname, then ask how to help.

✅ Figure: Good example - One line in the global instructions file turns every session start into a small surprise

Codex, GitHub Copilot CLI and other agents read their own global files, so the same line works there too. The global instructions rule lists where each agent looks.

Let the agent pick something fitting

The agent knows which repository it's in, so the nicknames often fit the work. Here's a session opened in a vertical slice architecture repository:

✅ Figure: Good example - The agent greets you with a nickname that fits the repository

And a few more from different sessions:

Figure: Example greetings. No two sessions are the same

Use the greeting as a canary

The greeting is also a free health check. Your global file holds your important rules: how you want code written, what to verify, what never to do. If the agent loaded the file, its first reply includes a nickname. If you get a plain "How can I help?", something is probably wrong. Open with a quick "hi" so you see the greeting before you hand the agent real work, and check the file if the nickname is missing.

Common reasons the greeting goes missing:

• Different tool - You opened a different agent, and it reads a different global file
• Different machine - Cloud agents, CI runners and remote containers don't have your home folder
• Broken file - The file moved, a symlink broke, or a bad edit left it unreadable
• Stripped context - Some modes and flags skip user-level instructions on purpose

A canary only works when it's cheap and you notice it straight away. A one-line greeting is both.

Keep it light

• Keep it short - The global file loads in every session, so one line is plenty
• Greet once - Ask for a greeting at the start of the session, not on every reply
• Keep it kind - Funny, not insulting. You read it before your first coffee
• Keep it out of automation - Subagents and headless runs on your machine load the same global file. Limit the greeting to interactive sessions so it doesn't end up in scripted output
• Keep it yours - Your nickname belongs in your global file. Use a personal, git-ignored project file such as CLAUDE.local.md if it should only happen in one repository