guides
Field notes from the trenches. How we build vaults, wire up AI agents, and ship systems that actually run in production.
A complete walkthrough of the Virtuoso vault system — conventions, frontmatter, folder structure, AI-friendly formatting, and the CLAUDE.md pattern that makes your notes AI-parseable from day one.
Production patterns for running a small constellation of services on one server. Nginx vhosts, systemd units, Let's Encrypt, deploy workflows, and when to stop.
Production patterns for using specialized AI agents in internal security monitoring and external auditing. Real workflows, verification discipline, and re-entry ramp for practitioners.
Why monolithic system prompts fail, and the framework-agnostic split that fixes them: SOUL (identity), AGENTS (behavior), HEARTBEAT (proactivity) — plus USER and MEMORY.
What a skill actually is, when to reach for one versus an MCP server or a subagent, and how to write skills Claude will actually invoke. Opinionated lessons from shipping 80+.
Install, configure, and live with a multi-agent system. Covers openclaw.json anatomy, agent file architecture, model tiers, plugins, and the failure modes you'll hit at 2am.
End-to-end tutorial for building a specialized AI agent with OpenClaw. Covers the SOUL + USER + HEARTBEAT pattern, workspace structure, cron-driven agent loops, and how to adapt this for any domain.