The seven-day path puts a useful sequence here: first make the assistant less generic, then give it access to the digital surfaces where value appears quickly.
Day 3 changes the assistant from generic to personal
OpenClaw 101 uses three files as the core of personalization: SOUL.md for tone and behavioral rules, USER.md for who you are and what you care about, and AGENTS.md for work habits and boundaries.
- SOUL.md answers how the assistant should speak and what it should never do.
- USER.md gives it your role, projects, preferences, and current priorities.
- AGENTS.md is where you tighten quiet hours, memory rules, and group behavior.
Do not over-design those files on day one
Day 3 makes a good point: these files are meant to be grown, not perfected. Write a basic version now, then keep editing whenever you notice 'it should have done this differently.'
Day 4 is the real shift from toy to tool
Once the assistant starts touching Gmail, Google Calendar, search, and browser access, the interaction changes. It stops being a personality experiment and starts becoming a helper that can actually check, fetch, summarize, and remind.
The fastest first task to aim for
- Ask it to review your new mail and identify what is actually important.
- Ask it what tomorrow looks like on your calendar.
- Ask it to search one current topic and summarize the result.
- Ask it to open one page and tell you what is on it.
What this step is really trying to prove
If the assistant already sounds more like your assistant and can now look at mail, calendar, or the web on your behalf, you have reached the first practical milestone. It no longer only answers. It helps.
Sources
- OpenClaw 101路Third-party路Community-curatedOpenClaw 101 Day 3
- OpenClaw 101路Third-party路Community-curatedOpenClaw 101 Day 4