governed-agents

What problem does it solve?

Consolidates verification and reputation management to prevent hallucinated success by AI sub-agents. It organizes a deterministic 4-gate coding verification suite and a 3-layer verification pipeline for open-ended tasks.

Core Features & Use Cases

• Deterministic verification gates for coding tasks (files, tests, lint, AST) ensure deliverables exist and meet quality checks.
• Open-ended task governance via Structural Gate → Grounding Gate → LLM Council with an EMA-based reputation system.
• Persistent task history and reputation tracking across tasks for progressive supervision and risk-aware task allocation.

Quick Start

Spawn a governed sub-agent for a coding task or open-ended task and observe automatic gating, council verification, and reputation updates.