RFC-ATF-11 specifies the Contract Formation and Convergence Layers of the Agent Trust Fabric — the eleventh RFC in the ATF Open Standard series published by OMNIX QUANTUM LTD. RFC-ATF-11 closes two structural gaps that remain after RFC-ATF-10: The Pre-Execution Void (IPFL): The ATF governance chain through RFC-ATF-10 contains no single PQC-sealed artifact that constitutes the complete governance contract before any agent turn fires. The DR, SAR, and MBR are separately signed but not jointly sealed. There is no formal declaration of counterparties (CPS) or freshness bounds (FPS). The Intake and Predicate Formation Layer (IPFL) closes this gap with the Governance Contract Formation Record (GCFR) — the world's first formally specified, five-predicate, PQC-sealed pre-execution governance contract for AI agent decisions. The GCFR seals five independently-signed predicates: IAD (agent identity + authority budget), SAR (scope + risk level), MFR (mandate reference), CPS (counterparty declarations), FPS (freshness bounds). Seal formula: SHA3-256 (iadₕash || sarₕash || mfrₕash || cpsₕash || fpsₕash) — normative and immutable (IPFL-INV-007). The Aggregation Heuristic Problem (GDCL): When multiple independent RetroactiveSemanticAssessments (RSAs, RFC-ATF-4) from heterogeneous governance domains must converge into a single decision, every implementation invents its own heuristic — destroying cross-implementation interoperability. Canonical case (articulated by Peter Cranstone, 3PMobile): a pharmaceutical AI decision must simultaneously satisfy pharmacy, payer, emergency, claims, and audit governance domains. What is the composite verdict? The Governance Decision Convergence Layer (GDCL) closes this gap with a deterministic worst-case-priority algorithm and a seven-verdict taxonomy (GDCL-1 through GDCL-7) producing a GDCLConvergenceRecord (GCR). IPFL architecture: Five intake predicates: IAD (Intake Authority Declaration), SAR (Scope Authorization Record), MFR (Mandate Formation Record), CPS (Counterparty Predicate Set), FPS (Freshness Predicate Set). Each independently PQC-signed (ML-DSA-65). GCFR: Assembles five predicate hashes into a single intakeₛeal. Any predicate modification invalidates the seal. Formed before Turn 0 (IPFL-INV-001). Append-only (IPFL-INV-008). Seal verified on every read (IPFL-INV-003). FPS freshness enforcement: Session rejected if now > fps. governanceᵥalidᵤntil (IPFL-INV-005). IAD-DR cross-reference: authoritybudget consistency check between IAD and DelegationReceipt (IPFL-INV-006). GDCL architecture: Seven-verdict taxonomy: GDCL-1 FULLYCONVERGENT (all PORTABLE), GDCL-2 CONVERGENTWITHDRIFT (no CRITICAL/INCOMPATIBLE), GDCL-3 CRITICALDRIFTDETECTED (≥1 CRITICAL), GDCL-4 PARTIALINCOMPATIBILITY (≥1 INCOMPATIBLE, not all), GDCL-5 FULLINCOMPATIBILITY (all INCOMPATIBLE), GDCL-6 INDETERMINATE (<2 RSAs), GDCL-7 CONVERGENCEERROR (schema violation). Worst-case priority (GDCL-INV-002): Composite verdict at least as restrictive as the most restrictive RSA. A single DRIFTCRITICAL in a set of 10 PORTABLE RSAs produces GDCL-3. Deterministic (GDCL-INV-003): Identical inputs always produce identical output regardless of RSA submission order. RSA hash verification (GDCL-INV-004): Invalid RSA contentₕash → GDCL-7. 14 new invariants are introduced: IPFL-INV-001–008 and GDCL-INV-001–006. Combined with the 159 invariants of RFC-ATF-1 through RFC-ATF-10, the ATF stack reaches 173 formally specified invariants across 29 protocol families. An implementation complying with RFC-ATF-1 through RFC-ATF-11 is designated ATF-IPFL-Compliant — the eleventh compliance tier in the ATF stack. Regulatory alignment: EU AI Act Art. 9 (GCFR = pre-execution risk management documentation), Art. 12 (five-predicate GCFR = automated logging of governance parameters), Annex IV (GCFR maps to technical documentation requirements) ; NIST AI RMF GOVERN 1. 7 + MAP 3. 5; ISO/IEC 42001 §6. 1. 2 + §9. 1. Related ADRs: ADR-204 (IPFL/RTE-001), ADR-206 (GDCL). RTE-001 verifyᵢntake: 36 IPFL-specific checks. 187/187 total PASS.
Harold Alberto Nunes Rodelo (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: