WIKI_INDEX

Integrity Protocol Wiki — Index

Content catalog. Every page represents something that actually exists in the codebase right now — see the schema's "no aspirational content" rule. Last updated: 2026-07-12 | Total pages: 25 (17 concepts, 8 entities)

Acronym glossary

• AIS — Agent Integrity Score
• BAA — Business Associate Agreement
• BCC — Behavioral Commitment Chain
• DID — Decentralized Identifier
• VTL — (old term) Verifiable Trust Layer — see Observability & PHI Safety for what's actually built
• ZKP — Zero-Knowledge Proof(ing pipeline)

Concepts

• Agent Primitives (Self-Sovereign Identity) — the 7 per-agent contracts; start here
• ComplianceGate & Xibalba Shield — the HIPAA/healthcare vertical
• Agent Integrity Score
• Behavioral Commitment Chain
• Merkle Batching & Anchoring Convention
• Decentralized Identifier
• Zero-Knowledge Proving Pipeline
• Integrity Market — prediction markets, binary options, A2A capital allocation (built, live on Base Sepolia)
• Smart BAA — on-chain Business Associate Agreement escrow (built)
• Local Metrology — client-side AIS signal derivation in the SDK (built)
• Observability & PHI Safety — the Redactor (built) + LLM-as-judge design ([PLANNED])
• Identity Ceiling & Verification Ladder — [PARTIALLY BUILT]: tier is server-verified + consulted by bcc_middleware's OPA gate for clinical intents; the AIS ceiling clamp itself is still [PLANNED]
• Cross-Chain Reputation Sync — [PLANNED]
• A2A Negotiation Protocol — [PLANNED]
• ZK-ML Model-Inference Verification — [PLANNED]
• Testing Strategy — the 3-layer test pyramid, incl. new Playwright E2E (built)
• AIS API — Versioned Wire Spec — the generated, externally-supported /v1/* spec at spec/ais-api/ (built)

Entities (built)

• contracts — Solidity/Foundry: the 7 primitives, factory, registries, XNS, Shield, market layer, $ITK, reworked CCIPReputationBridge (165 tests, live on Base Sepolia — XNS/CCIP not yet broadcast)
• integrity-oracle — Rust/Axum AIS scoring + on-chain verification + markets/leaderboard/wallet reads + PHI-rejection backstop, ASCII-escaping canonical-JSON fix (54 tests + e2e)
• integrity-sdk — Python agent library: registration, BCC, markets, telemetry, PHI redaction, new pre-execution intent-capture (invoke_intent), fixed telemetry-signing wire bug (97 tests + 1 opt-in oracle e2e)
• integrity-cli — developer CLI, real on-chain register incl. real oracle re-verification, new xns command group (56 tests + 1 opt-in oracle e2e)
• bcc_middleware — FastAPI + OPA policy gate, incl. verification-tier gate (54 + 16 OPA tests)
• integrity-mvp — the dashboard app, rewritten (2026-07-12) into a new 16-page shell: real oracle/userapi reads and writes throughout (agents, markets, wallet, Smart BAA sign/revoke, real enterPosition bet placement, real userapi account/API-key auth), a Notion-style Block & Widget dashboard engine, a rewritten "Verify Agent Control" flow (the old signature-challenge "claim" premise had no on-chain support), real test infra (9 vitest + 18 Playwright e2e, run against a live backend+chain), and real project docs. Explicitly seeded (badged, not silently faked): order-book UI, node/network telemetry widgets, BAA creation.
• integrity-zkp — real Noir/Barretenberg circuit, compiled & proven
• integrity-userapi — FastAPI + Postgres user accounts/auth, strictly non-chain (33 tests, real Postgres, real CORS for integrity-mvp)

Guides

• Smart Contract Development — how to write, test, and deploy a new contract in contracts/: repo conventions (AccessControl, custom errors, NatSpec), Foundry test patterns (vm.prank/makeAddr/vm.expectRevert(...selector)), wiring into Deploy.s.sol/DeployMarkets.s.sol + make sync-abis, local/Base Sepolia deploy walkthrough, and the SovereignAgent.execute vs. direct-EOA auth convention.
• Multi-Domain Guardrails Design — [DESIGN, PARTIALLY BUILT]: how bcc_middleware's HIPAA-only OPA policy gate generalizes to pluggable, domain-scoped guardrail bundles (industry survey + concrete design against this repo's existing bcc.rego/HIPAAGuardrailRegistry.sol pattern). A real worked-example bundle (bcc_middleware/policies/general.rego, 12 passing OPA tests) is built as a stretch goal; the domain-aware GuardrailRegistry.sol and bcc_middleware wiring are still design-only.

Architecture

(none yet — add a cross-package data-flow doc if the per-package entity pages prove insufficient)

Open queries

• No LLM-as-judge rubric exists anywhere in this repo or the cross-checked Desktop spec docs ("Xibalba Solutions defines" it, per the plan notes) — the judge_evaluations ingestion schema is designed but the actual scoring rubric is an open product question, not an engineering one. See Observability & PHI Safety.