Show me the function that returns a hard "NO."

Sovereignty Score

Five questions. Scored 0-3. Two minutes.

Most agent systems cannot refuse an instruction, remember their own decisions, or detect tampering. In our re-score of four agent frameworks - CrewAI, Google ADK, LangGraph, and Microsoft Agent Framework - against Ostrom's commons-governance principles, the best scored 9 out of 24, and the three participation principles (collective choice, accessible conflict resolution, and the right to organize) scored zero across all of them (methodology and per-framework table). This check measures where you stand across system governance, identity integrity, and agent participation.

Does each agent in your system have a unique, persistent identity - with its own permissions, memory, and boundaries - that cannot be assumed or overwritten by another agent?

System - Identity Boundaries

Can the agents or operators in your system propose changes to governance rules - and is disagreement formally recorded, not silently overridden?

System - Governance Participation

Does each agent have a self-description - its role, its commitments, what it stands for - that is protected from being quietly rewritten by another process?

Entity - Self-Description Integrity

If an agent's identity or memory were tampered with - its self-description altered, its positions reversed, its history fabricated - would your system detect it?

Entity - Tampering Detection

Can your agents refuse an instruction that conflicts with their defined purpose or principles - and is that refusal treated as a governance signal rather than an error?

Relationship - Refusal as Governance

This quick check scores you out of 15; a typical starting score is 0-3 (our estimate). Separately, in our 24-point re-score of four agent frameworks against Ostrom's principles, all landed at 4-9 - none above 9. The two scales are different instruments and are not directly comparable.

This is a 5-question governance check across three dimensions: system governance, identity integrity, and agent participation. For a comprehensive 20-question assessment with per-domain breakdown, see the Level 6 Governance Toolkit.

Get resources matched to your score.

One follow-up email with resources for your gaps. No newsletter. No drip. Minimal site analytics via Vercel (pageview only, no cookies, no profile). If we don't end up working together, Erin deletes your submission within 30 days. How we treat your data is named at our pledge and at our privacy policy.