Evidential Protocol

Matsés + Kusunda inspired epistemic metadata for AI agent systems. Every claim declares how it knows.

Subsystems

• [types] — Core EP/1.0 types (EvidenceClass, Evidence, TrustScore, etc.)
• [schema] — JSON schema generation and envelope formatting
• [validate] — Validation rules for claims and evidence
• [audit] — Bilateral audit ledger with indexed queries
• [provenance] — Root provenance tracing through claim chains
• [fingerprint] — Behavioral fingerprinting and anomaly detection
• [ergative] — Role-based permission gating for tools
• [attestation] — SHA-256 content attestation and Solana memo formatting
• [scoring] — Continuous multi-axis scoring engine