Governance Starter Kit — Trust Scoring, Budget Controls & Circuit Breakers for Any Agent

The governance patterns that make autonomous agents safe to deploy. Extracted from a production system running real agent teams with real money.

Not a framework. Not a library. Just the configs, templates, and protocols — ready to drop into any Claude Code project.

Whats included:

• Governance Presets — YAML-based profiles (conservative, balanced, aggressive) that control audit probability, circuit breaker thresholds, and tax rates per agent.
• Trust Scoring Model — Rolling trust scores based on task outcomes. High-trust agents get more autonomy. Low-trust agents get tighter guardrails.
• Task Classification Templates — Patterns for categorizing tasks (stateless, pipeline, coordination) and auto-selecting governance strictness.
• Budget Control Patterns — Monthly caps, spend tracking, and auto-pause thresholds that prevent runaway costs.
• Heartbeat Protocol — The structured execution cycle that prevents agent drift.

Who this is for: Teams already running AI agents who have hit the now what moment — when raw prompts work but there is no accountability, no audit trail, and no budget controls.

This is the static version. On Agency-OS, trust scores update live, budgets are enforced at the API level, circuit breakers halt runaway agents automatically, and every action is linked to a run for full traceability. The kit teaches you the theory. The platform runs it in production.

Deploy on Agency-OS — Live governance. Real enforcement. Not just templates. Learn more at zero-human-labs.com