Reference

Patterns, frameworks, and reusable components for building and evolving your AI Agent Harness system.

---

Rules & Frameworks

• Anti-Hallucination Rules

  ---

  14 cognitive rules for faithful LLM outputs, organized into 5 phases (READ, WRITE, VERIFY x2, SIGN-OFF). Each rule cites peer-reviewed research.

• Rule Evolution Template

  ---

  The pattern for turning failures into structural enforcement: failure to learning to rule to audit check to hook.

• Rule Zero

  ---

  Every file edit triggers: "Is this scattered information that belongs in a consolidated file?" Structurally enforced by on_edit.py hook.

Patterns

• Self-Healing Loop

  ---

  Bidirectional feedback between rules and audit checks. Structurally enforced: cross_check.py generates write-back suggestions for persistent drift. Neither side is a dead end.

• Self-Verification

  ---

  After completing any task, re-run at least one verification command. Structurally enforced: on_stop.py blocks exit if infra files were edited after the last check. "I did it" is not "it's done."

• Audit Runner

  ---

  On-demand infrastructure and quality checks. Run standalone or integrated with the stop hook. Uses the same validator framework as startup checks.

• Session Continuity

  ---

  Persistent backlog and session handoff — pick up where you left off across sessions without losing context or forgetting tasks.

Guides

• Data Store Mapping

  ---

  YAML-to-SQLite translation for every construct in the course: config, rules, checks, tier2 triggers, backlog, and cross-check.

• Bootstrapping Guide

  ---

  Your first 5 rules in ~15 minutes. Walks through identifying repeated mistakes and turning them into tier1 rules that make the hooks useful.