AgentLoom: Executable Governance Framework for Builder Agents

AgentLoom v3 is a runnable framework for knowledge-graph-grounded agent governance: propose-review-accept protocol, eight Tier-A executable validators, JSON-schema KG integrity checks, and a read-only human review dashboard. Builder and domain roles share one governance model; domain knowledge accumulates through audited proposals rather than ad-hoc prompt edits. Dual-helix architecture (learning + governance) positions the framework for trustworthy agentic AI in production settings.