zurbrick

19 production-hardened patterns for AI agents covering memory, verification, ambiguity handling, and self-improvement. Use these patterns to enhance agent reliability, prevent hallucinations or context loss, ensure seamless handoffs, and address silent delegation failures in production.

Practical memory architecture for OpenClaw agents. Use when an agent needs to remember commitments across sessions, retrieve prior context, prevent repeat mistakes, stage durable-memory updates safely, compile reusable procedures, or improve memory hygiene over time. Supports a 3-tier adoption path: Core, Recommended, and Advanced.