# E24 — IP Candidates (engineering inventory for counsel)

**No patentability, novelty or valuation determination is made here.** Prior-art search and qualified
counsel are required. This is a factual map of mechanisms that exist, are tested, and are evidenced.

| ID | Candidate mechanism | Mechanism (what/why) | Implementation evidence | Claim surface (illustrative) | Prior-art questions |
|---|---|---|---|---|---|
| C-01 | State-bound protected-execution authorization | protected execution requires a binding to the current committed governed-state root; stale/rolled-back state invalidates authorization | `cain/e22/authbinding.py`, `E22_GATE_RESULTS.json` | method for binding authorization validity to a state root with staleness abort | time-of-check/time-of-use authorization binding prior art |
| C-02 | Single-use, state-bound, scope-limited execution permit | permit binds principal/agent/tenant/capability/target/state-root/nonce/expiry; atomic single-use consumption | `cain/e22/permits.py`, `tests/e22` | method for atomically-consuming a state-bound capability permit | capability/lease token prior art |
| C-03 | Compositional authority-conservation engine | `AUTHORITY_OUT ⊆ AUTHORITY_IN` checked per transition and across a chain, with an explicit authority-delta record answering 9 questions | `cain/e23/authority.py`, `cain/e24/authority.py`, `E24_SUBSTRATE_PROOF.json` | method for enforcing authority conservation across a multi-product pipeline | capability attenuation prior art |
| C-04 | Governance lineage graph with orphan/expansion detection | machine-validated AGENT→…→STATE_UPDATE chain; detects missing predecessor, authority expansion, orphan execution/evidence, tenant mismatch | `cain/e23/lineage.py`, `E23_COMPOSITION_RESULTS.json` | method for validating a governance execution lineage | provenance-graph prior art |
| C-05 | Live shadow divergence engine for authorization-enforcement migration | observes the real authorization path and classifies E8↔E22 divergence without changing the decision | `cain/e23/shadow.py`, live `GET /frontier/e23/shadow` | method for safe promotion of an enforcement dependency under observation | shadow/deferred-decision prior art |
| C-06 | Deterministic replay envelope separating logical/cryptographic/external determinism | replay reproduces governance logic and records random ids/timestamps explicitly instead of deleting them | `cain/e23/replay.py`, `E23_FRONTIER_PROOF.json` | method for classifying nondeterminism in governance replay | deterministic-replay prior art |
| C-07 | Frontier substrate envelope + domain-separated state dimensions | one attenuation-only envelope; 12 domain-separated state dimensions so one product cannot substitute another's state | `cain/e24/substrate.py`, `E24_SUBSTRATE_PROOF.json` | method for a multi-tenant product substrate with non-substitutable state commitments | envelope/context propagation prior art |
| C-08 | Machine-validated composition graph rejecting authority-granting edges | typed edges; rejects observation/evidence/reputation/prediction→authority and product→execution bypass | `cain/e24/composition_graph.py`, `E24_GRAPH_RESULTS.json` | method for statically validating a product composition architecture | data-flow policy prior art |
| C-09 | Epistemic risk fusion that refuses to average observed/simulated/predicted into authority | segregates epistemics; contradictions flagged, never averaged; never grants authority | `cain/e24/risk.py`, `tests/e24` | method for governed multi-source risk fusion | ensemble/risk-fusion prior art |
| C-10 | Zero-CAIN-import clean-room verification of composed proofs | independent recomputation of state/authority/lineage/certificate across E21–E24 | `verification/e2{1,2,3,4}/cleanroom/*` | method for independent verification of a governance proof bundle | supply-chain attestation prior art |

## Trade-secret potential
The accumulation of negative/attack knowledge (the invariant families I13–I108, the chaos/Byzantine
corpora, and the clean-room verification discipline) is a trade-secret-style asset that is genuinely
hard to reproduce even with the source, because it encodes *what was tried and refused*, not just the
final code.

## Dependency on public disclosure
The E23/E24 bundles are published for verifiability; any patent strategy must account for that
disclosure date. Counsel should assess.
