If you're still using a keyboard to communicate with your agents, you're wasting valuable time! Here's why it's time to grab a mic, teach your agent some basic skills, and bring them into your world instead of trapping them behind a screen.
The era of asking chatbots questions and then patiently waiting for them to respond is over. The only secret to truly leveling up your productivity is making your agents proactive, so they're taking independent actions when they're most essential, rather than waiting around for a prompt.
The best way to do that? Opening up a direct vocal line of communication between you and your agent swarm. You need to give them instructions as you come up with brilliant ideas -- no matter where in the world you happen to be. And they need to alert you about important updates, even when you're away from your work station.
STEP 1: INTRODUCE YOURSELF
OpenClaw agents are already pretty smart, right out of the box. But they are also generic: a blank slate with no personality, identity, or goals. Fortunately, it's easy to bring your agent to life, and start them on their journey toward becoming a true virtual assistant, rather than just another bot.
The two most important files for these purposes are:
SOUL.md and USER.md
SOUL.md is just what it sounds like. The home of your agent's personality, identity, and perhaps most importantly, their tone. Looking to create a just-the-facts Type A fixer who gives it to you straight? Or are you building a companion who will hang out with you while you get your work done? This is where you tell them who you're looking for.
USER.md provides the agent with information about YOU, the user. It gives them valuable context about your world, your contacts, your company, and the sorts of projects on which you're hoping to collaborate.
STEP 2: TALK IT OUT
Now you COULD just update these files yourself, directly. They're located in your workspace folder, and you can change them with any text editor:
nano ~/.openclaw/workspace/SOUL.md
nano ~/.openclaw/workspace/AGENTS.md
You can also give your OpenClaw agent instructions in plain language, and let them update the files on their own. for example.