How to turn your OpenClaw into the world's best assistant

I turned my @openclaw into the most effective assistant and chief of staff I've ever worked with. I've hired EAs in my previous companies and I'm absolutely blown away by how well this is working. This post is a step-by-step guide to how I did it. (If you want to cheat, you can just tell your OpenClaw to follow this open source repo exactly: github.com/snarktank/clawchief) Now let me show you how it works. What can it do? Schedule all my meetings like "Hey R2, can you please find a time that works for both of us?" (he can even parse Calendly links and pick times that work for me and book them). Check my inbox every 15 minutes and tell me which emails actually need my attention Proactively follow up on emails that didn't get a reply Watch my calendar, flag conflicts, and warn me when I have something coming up soon Run my day from one canonical markdown task list in `tasks/current.md` Prep my task list before I wake up by promoting due items into `Today` Keep my task list clean by avoiding duplicate tasks across backlog and today views Spot replies from prospects or referral partners and update the outreach tracker in Google Sheets Update my CRM when I send/receive emails Research suppliers and reach out to them to setup meetings Send me short, high-signal updates when something needs action, and stay quiet when nothing important is happening Work from durable context in files, memory, Gmail, Calendar, and Sheets instead of pretending chat history is enough Be customized around my business, my channels, my preferences, and my exact operating style instead of acting like a generic AI assistant clawchief includes: Step 1: Start with a working OpenClaw install Before you do anything with clawchief, make sure OpenClaw itself is already installed and working. clawchief is not a replacement for OpenClaw. It's a practical operating layer on top of it. Step 2: Get GOG working first This setup expects gog to work for: Gmail message search Calendar list and event reads Google Sheets metadata reads If those are broken, your assistant won't be able to do real executive-assistant work reliably. Talk to your claw about how to set it up and they'll help you. Step 3: Install the skills Copy these skill directories into ~/.openclaw/skills/ skills/executive-assistant skills/business-development skills/daily-task-manager skills/daily-task-prep These are the behavioral building blocks. This is what teaches OpenClaw how to act like an executive assistant, how to manage a task list, and how to handle operational business-development workflows. Step 4: Install the workspace files Copy these into ~/.openclaw/workspace/ workspace/HEARTBEAT.md workspace/TOOLS.md workspace/tasks/current.md If you already have files there, merge carefully. Why HEARTBEAT.md matters This defines what the assistant should check proactively. For me, that includes things like: important new email upcoming calendar events scheduling issues task-list follow-up occasional marketing or relationship-building nudges This is how you stop your assistant from being passive. Why TOOLS.md matters This is where I keep local, environment-specific notes. Not generic skill logic — actual setup details. For example: preferred email accounts Google Sheets usage notes local environment quirks browser-profile guidance small tactical rules I don't want buried in prompts Why tasks/current.md matters This is one of the most important files in the whole system. I keep one canonical markdown task list. That means when the assistant checks what matters today, it is looking at one live source of truth instead of guessing from stale conversation history. Step 5: Create your private context files Make sure these are highly customized to you and your OpenClaw: AGENTS.md SOUL.md USER.md IDENTITY.md MEMORY.md memory/ This is where OpenClaw becomes your assistant instead of mine. These files let you define: who the human is who the assistant is tone and boundaries personal and business preferences long-term memory continuity across sessions If you skip this step, you'll have a decent template. If you do this step well, you'll have an assistant that feels personal, grounded, and increasingly excellent. Step 6: Replace every placeholder The repo includes placeholders for the obvious things: owner name assistant name assistant email primary work email personal email business name business URL timezone primary update channel primary update target Google Sheet ID target market target geography repo root Step 7: Set up cron jobs This is where the assistant starts to feel alive. The repo includes a cron template. The recommended starting jobs are: executive assistant sweep daily task prep daily business-development sourcing You can add optional jobs later, like backups or self-update, but I would start with the core operational routines first. The important point is this: The assistant becomes dramatically more useful when it wakes itself up to do recurring work. That is what shifts it from r