Verticity and Hazard-Agent Encoding: A Unified Proof Architecture with Capsule-Grade Instantiation

This deposit formalizes Verticity as a structural property of bounded-recursive coordination systems: when shock increases interpretive branching faster than recursion can reconcile it, approximately optimal governance collapses toward axis-threshold rules on low-dimensional summaries. The paper is instrument-first and referee-auditable: it enforces a claim boundary (no myth literalism; no universality beyond estimator class), ledger separation (Λ/E/A/Σ), an interpreter lock (“No STDOUT = No Claim”), explicit falsifiers, anti-congruent discriminators, and fork-discipline requirements. A bridge result (“Hazard-Agent Encoding Lemma”) yields a falsifiable prediction for spatially fixed recurrent hazards with catastrophic tails: agentic encoding should carry higher explicit hazard motif load than unanchored carriers and place-matched non-agent mentions under declared null models. A capsule-grade certified instantiation is published separately as Dragons as Hazard Agents (v1.1): https://doi.org/10.5281/zenodo.18740729. The companion record includes a provenance wall (Ledger-E witness excerpts), a frozen analysis contract (including a predeclared site-null module), a machine-executable audited analysis table (CSV), embedded artefacts (results/manifest/certificate), and a verifier script. This Verticity deposit imports the empirical claim by DOI reference and makes no standalone numeric claims.