banking-agent-audit: Governance Patterns for Autonomous AI Agents in Regulated Banking

Reference governance patterns, schemas, and runnable implementations for audit-ready autonomous AI agents in regulated banking. Ships five corrected governance primitives (an autonomy-ladder level-gate, a fail-closed un-self-clearable sovereign veto, a tamper-evident hash-chain ledger with witness-anchored regeneration detection, a DEFCON transition-direction guard, and an effective-challenge harness) plus two implemented and tested banking controls (model-risk validation gated on attested challenger independence, and an ECOA/Regulation B section 1002.9 adverse-action gate with the FCRA section 615 overlay) and a sanctions/AML disposition-workflow pattern that bundles no sanctions list and screens against a pluggable, deployer-wired list provider. Reference IP for adoption, not a control operating in production; not legal or compliance advice.