Operations & Maintenance

Sync your OpenClaw code repo to GitHub every hour. This captures both your manual changes AND changes OpenClaw makes to itself.

OpenClaw is self-evolving — it modifies its own code, skills, and markdown files. An hourly cron job checks for any changes (yours or the agent's), commits, and pushes to GitHub. This is your primary rollback mechanism. If something breaks, you can revert to any hourly checkpoint. Keep databases out of git — they go to Google Drive instead.

Databases don't belong in git. Back them up to Google Drive on a schedule with timestamped snapshots so you can restore any point in time.

CRM data, analytics snapshots, knowledge base, business analysis, cron logs — all backed up to Google Drive with timestamps. Keep a detailed restore document that describes how to rebuild everything from scratch. Git handles code; Drive handles data. Together, they mean you can rebuild your entire system from zero if the MacBook Air dies.

Let OpenClaw's built-in memory system distill daily notes into long-term patterns and preferences automatically on a weekly cycle.

During each day, conversations, completed tasks, and mistakes all get piped into daily notes. Weekly, the system synthesizes these notes: distilling patterns and preferences into long-term memory, and extracting corrective patterns and mistakes-not-to-repeat into a learnings folder. Over time, the system gets better without any manual curation.

Have OpenClaw download platform best practices and the model's prompting guide, then cross-reference all your markdown files daily to catch drift and anti-patterns.

Markdown files drift as you add skills and tell OpenClaw new things. Once a day, the system downloads OpenClaw's official best practices and the current model's prompting guide (e.g., Opus prompting guide from Anthropic), stores them locally, and cross-references every markdown file against both. It flags anything that contradicts best practices or uses deprecated patterns and recommends changes.