Machine Observer separates identity, authority, reputation and validation rather than collapsing them into one trust score. A cryptographically identifiable agent can still be unreliable; a popular agent can still be unauthorized.
The 0–100 score measures available trust mechanisms. It does not certify vendors, agents or identities.
Proposes per-agent keys, signed outbound actions and policy enforcement rather than agents inheriting unrestricted user credentials.
Identity 20/20 · authority 15/15 · validation 5/10 · interoperability 15/15Binds agents to verified human or business principals and issues signed KYA/KYA+Pay credentials.
Identity 13/20 · authority 10/15 · validation 10/10 · interoperability 13/15Defines separate identity, reputation and validation registries for cross-organizational agent trust.
Identity 17/20 · authority 3/15 · validation 10/10 · interoperability 15/15Purpose-built agent identities, blueprints, sponsors, Conditional Access, risk signals and lifecycle governance.
Identity 10/20 · authority 15/15 · validation 10/10 · interoperability 8/15Provides strongly attested cryptographic workload identity and federation across trust domains.
Identity 17/20 · authority 10/15 · validation 10/10 · interoperability 13/15General-purpose machine-verifiable credential substrate rather than an agent-specific identity system.
Identity 17/20 · authority 5/15 · validation 10/10 · interoperability 15/15Standardizes Agent Cards and authentication schemes for cross-framework agent communication.
Identity 13/20 · authority 8/15 · validation 0/10 · interoperability 15/15Authentication proves a presented credential under some trust root; it does not prove good behavior. Reputation is not authority. On-chain registration is not safety certification. Enterprise policy is not public portability. Vendor assertions remain labeled vendor evidence, and drafts remain drafts.
Popularity, funding, advertising and Machine Observer ownership do not improve scores.
Machine Observer’s own MO-ENG role manifest is listed as a related first-party implementation but receives no score. The goal is to test role identity ideas without rewarding ourselves for inventing the test.