cryptographic readiness

This track retains the earlier Hemera/Eidos objectives and supplies exact profiles to the A–C integration work. Local development can exercise versioned experimental profiles; production acceptance requires the evidence below.

H1: finalize Hemera's primitive

Owner: Hemera.

• Resolve the profile: hemera 0.4 (x⁷ every round, RF=12, RP=48, Poseidon2 reference matrices) frozen 2026-10-16; the mixed x^7/inverse RF=8/RP=16 candidate (0.3) is the record. Open: the campaign's collision, preimage, second-preimage and transcript-use requirements (hemera/audit/campaign.md).
• Validate attack candidates against the real implementation; distinguish complete-profile results from local S-box properties and reduced models.
• If a requirement fails, carry a justified replacement through matched cost/security evaluation and production parameter selection.
• Freeze field, matrices, constants, rounds, domain separation, output/encoding and streaming behavior with versioned vectors and migration rules.

Evidence starts in Hemera research. The public campaign that gates the private pool (G3), with the margins of the shipped and proposed profiles and the bounty, is hemera cryptanalysis. The earlier local witness and degree findings retain their stated scope.

H2: structural commitments and the .cyb adapter

Owner: Hemera; format semantics supplied by the .cyb owner.

• Finalize the permutation/sponge → universal structural commitments → format adapter layering and its canonical blob/sequence/record encodings.
• Specify semantic section/chunk extraction, order, names, offsets and parser rejection behavior. Bind structure and domain unambiguously.
• Validate inclusion/opening rules, streaming/chunk equivalence and identity stability; version any change to existing particle/tree interpretation.

Contract: structural commitments.

E1: harden Eidos and refine the implementation

Owners: Eidos kernel/checker; Hemera theorem and implementation models.

• Validate inductive environments, normalization/resource bounds and the closed-term checking boundary; reject unsupported claims explicitly.
• Establish universal field and machine-word arithmetic obligations, including noncanonical inputs, overflow and reduce128, in checked proof terms.
• Prove inversion-chain, encoding, padding, permutation, streaming and tree refinements against the actual implementation and its pinned sources.
• Maintain negative controls and source/model conformance. Report cryptanalytic assumptions separately from functional refinement and checker trust.

Evidence: Hemera proof coverage; interface: strict checking.

P1: network-usable execution profiles

Owners: Zheng/Lens/Nox, Joy adapter and Cyber admission.

• Pin supported statements, encodings, verifier parameters and disclosure.
• Complete authenticated state reads/transitions and any advertised secret, hiding or succinct proof mode; connect them to node-owned trusted context.
• Cover program/input/output/state binding, forged openings, malformed sizes, replay/freshness and expected-job matching with adversarial tests.
• Version source compilation semantics and its validation separately from machine execution proofs. Migrate all root/digest/codec consumers together.

Exit: each production operation selects a fully specified, implemented and validated profile; the exact primitive and implementation evidence supports the security and disclosure claims made for that operation.

discover all concepts