Trust Transfer Protocol — open trustworthiness primitives for autonomous systems, non-human identities, and runtime authority. Includes PCTR (Consequence Preview, Trust Routing, Receipts) for AI agent safety and execution authority.
This is a specialized cryptographic/trust protocol library, not a traditional MCP server. Tool definitions exist but lack rigor expected for agent-facing interfaces. 19 tools have descriptions (good), but most parameter schemas are minimal or missing type details. No input validation guidance, error recovery hints, or output schema documentation. Naming is domain-specific but not action-verb-driven (e.g., 'trustZone' vs 'classify_trust_zone'). Parameters often lack descriptions or constraints. No evidence of pagination, batching, or idempotent operation patterns. Security-critical tools (registerIssuer, submitReceipt, authorize) lack permission gates or audit trail documentation.
Authorizes an execution request via the runtime authority, returning decision, trust score, and receipt
Builds a signed execution receipt from authorization request and decision result
Validates a TTP protocol file for correctness
Computes a binding hash from subject, action, resource, paramsHash, and timestampBucket
Computes trust score with decay applied based on time elapsed
Creates a signed execution receipt from request and trust route decision
Creates Express middleware to verify TTP trust tokens on incoming requests
Evaluates a TTP protocol file for a given subject at a specific time
Parameter schemas lack type definitions and constraints. Most parameters show only 'type' and 'description' with no enum, pattern, minLength, or range constraints. This invites invalid input from agents.
No output schema documentation. Tools return objects but the response structure is not formally documented. Agents cannot plan downstream calls or extract required fields reliably.
Security-critical tools (registerIssuer, registerAgent, submitReceipt, authorize) lack permission gates and audit trail documentation. No indication of required scopes or access control.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 45 | 2026-07-28+ | v2 |
Evaluates trust for a request considering risk, cost, compliance, and authority grant
Requests a trust token from the Trust Authority for a given domain
Registers an AI agent with the Trust Authority
Registers a trust issuer with the Trust Authority
Requests a trust token from the Trust Authority for an agent in a domain
Reroutes an action to a new trust route, ensuring the replacement route meets at least the same authority level as the previous route
Resolves a trust route through a graph of agents and actions, applying authority, trust, and consequence constraints
Resolves a trust route based on request, candidates, policy, and availability of resolver and proof engine
Submits a behavioral receipt to the Trust Authority for aggregation
Classifies trust score into a trust zone (active, degraded, warning, critical)
Verifies a TTP trust token signature and claims
Naming convention is not action-verb-driven. Tools use domain-specific names (trustZone, resolveTrustRoute) rather than verb_noun pattern (classify_trust_zone, resolve_trust_route). LLMs rely on verb prefixes to infer intent.
No error handling guidance. Tools do not document what errors can occur, whether they are retryable, or what the agent should do next. No recovery hints provided.