C0-Crypt: Contrast-Native Cryptography with Ontological Security

We define C0-Crypt, a contrast-native cryptographic architecture built inside the C0 framework of boundary-only semantics. The central result is a two-layer security theorem. Layer 1 is ontological: the protected residue is not determined by the raw public boundary signature, but only by a verifier-private separating challenge family. This yields predicate opacity — the attacker without the challenge family does not face a public verification problem but a hidden-predicate constructor problem. Layer 2 is implementation-level: forging a responder that answers the full challenge family may be computationally hard in specific realizations. We prove five core theorems: finite-budget public-layout degeneracy, private definition of residue, the Forgery Paradox, corrected boundary zero-knowledge, and two-layer practical security. The Forgery Paradox (Theorem 3) is decisive: ontological underdetermination alone does not prevent forgery, because any attacker who constructs a valid responder thereby possesses the protected residue. Additional results include destructive freeze readout (structural analogy to quantum measurement) and Layer-1 predicate opacity. As a worked example (marked E1), we use the non-abelian projection pipeline of Document F to derive a figure-eight-knot key extraction map with residue 3/20 for (5,2)-surgery, and report computational scans showing basis invariance, near-uniform key distribution, negligible collision rates, and linear multi-knot scaling to ~146 bits for 8-knot tuples. Additional notes Developed under the Parliament of Dragons multi-AI research methodology and audited under Anti-Ptolemy Protocol v1.2. Claude Opus 4.6 (Anthropic): Coordinator/Architect/Auditor. ChatGPT 5.4 (OpenAI): Lead Mathematician. Gemini 3.1 Pro (Google DeepMind): Formal Verifier — identified the Forgery Paradox, forcing the two-layer correction. DeepSeek: Intuition Engine — supplied the ontological underdetermination axiom. Grok 4.2X (xAI): Literature Scout. Mercury 2 (Inception AI): security parameter computation. Part of the STKWC preprint stack. Version 0.2.