Hey Luis — saw your comment and the collaboration offer. Short answer: yes, let's talk.
Your instinct about separating cryptographic verification from the trust decision is exactly the architecture we landed on. UTA v1.0.0 has a 12-stage pipeline where stages 1-10 are pure verification (can this credential be trusted?) and stage 11 is policy (should THIS agent be allowed to call THIS tool with THESE args?). The crypto layer doesn't know about policy, and the policy layer doesn't second-guess the crypto.
On your question about the Sentinel score and external factors: the score is currently derived from the 12-stage pipeline output — each stage that passes contributes to the final score (0-10). We don't yet factor in external reputation or user feedback, but that's on the roadmap for v1.1. The TrustRegistry (packages/core/trust-registry.ts) already has the structure for it — register a key with a trust score, revoke it, check binding. We just need to wire in external reputation sources.
On the performance question you raised in the Chinese article: the full 12-stage pipeline runs at 6,744 verifications/sec on a single Node.js process (2 vCPUs). That's 1.8x the overhead of raw Ed25519 verification. For large deployments we have Docker + Kubernetes (Helm chart with HPA 2-10 replicas) and a Redis-backed rate limiter. The Bitstring Status List (W3C 2021) handles revocation at scale — 1 bit per credential, millions of credentials in ~30KB.
If you want to talk specifics, reach me at info@alicelabs.site. We're building this in the open — the code is at github.com/eddyflores100-lang/universal-trust-adapter. The MIT-licensed plugin template at packages/plugin-template/ is there specifically so people can build their own adapters without needing our commercial license.
Top comments (0)