Repository navigation
feat(policy): require contribution, citation, and novelty review for open-problem exports #8
Description
Problem
A kernel-verified packet may still be misleading if its public presentation omits prior work or overstates novelty. MathCorpus needs a publication policy that treats correctness, provenance, attribution, and novelty as separate review dimensions.
Scope
Add publication-readiness rules for packets and proof variants related to open problems, new bounds, counterexamples, or claimed new methods.
Require:
- a contribution-class declaration:
new_proof,independent_rediscovery,formalization,verification,reconstruction,adaptation, orliterature_derived_synthesis - known-prior-art references
- literature-search scope and date
- citation-lineage review status
- novelty-review status
- AI-assistance disclosure
- unresolved attribution caveats
Suggested publication statuses:
metadata_onlykernel_verified_unreviewedcitation_review_pendingnovelty_review_pendingreviewed_with_caveatspublication_readyblocked_missing_attributionblocked_novelty_claim
Acceptance criteria
- Kernel verification alone cannot produce
publication_readyfor an open-problem claim. - Citation or novelty uncertainty never changes proof truth, only public claim status.
- Strong novelty language is blocked when prior-art review is incomplete or disputed.
- Exports include a human-readable and machine-readable contribution statement.
- Review records are reviewer-attributed, timestamped, and hash-bound.
- Policy supports later correction, revocation, or addition of citations without rewriting canonical packet history.
- Existing packets remain valid but are marked under the legacy/unreviewed policy until migrated.
- CDC serves as a fixture classified as formal verification/reconstruction with prior-flow literature disclosed.
Activity
Mnehmos commentedon Jul 12, 2026
Implemented (local commit
61de03e, stacked onae51810/928b2be/ed2e1a7/b3bfa63/bcdb683/8c74201/7f1ac10— not yet pushed, see notes on #1-#7). This is the last issue in the 8-issue MCIP roadmap.Purely a policy layer over #7's
contribution_statements/citation_reviews— no new record types needed.publicationobject onpacket.schema.jsonwith all 8 suggested statuses (metadata_only,kernel_verified_unreviewed,citation_review_pending,novelty_review_pending,reviewed_with_caveats,publication_ready,blocked_missing_attribution,blocked_novelty_claim).publication_ready:policy.check_publication_statusrejects it outright when the packet has zerocontribution_statements.publication_readyon anopen_problem_related: truepacket requires at least one current (supersedes-aware — a superseded review doesn't count)citation_reviewwithreview_status: "endorsed".contribution_class: "new_proof"declaration is rejected frompublication_readyunless a current review is endorsed — disputed or absent review blocks it.publication.statusis structurally separate fromtrust.rung/proof_authority/status; nothing in this repo lets one affect the other.citation_reviews[].supersedesnow has a dangling-reference check, so the correction/revocation chain actually validates end-to-end.policy.implicit_publication_statusgives every packet lackingpublicationa sensible default (kernel_verified_unreviewedormetadata_only) — no retroactive requirement, no migration needed.mathcorpus.export.contribution_summaryrenders a human-readable sentence (contribution class, known prior art, publication status, unresolved caveats), added to every exported row ascontribution_summary, alongside the machine-readablecontribution_statements/publicationfields already present.Real fixture, same honesty standard as #7:
union_closed_sharpness.v1getspublication.status: "reviewed_with_caveats"— deliberately notpublication_ready, even though it technically has the endorsed review needed to pass the gate, because its Yu-2023 prior-art citation carries an honestly-disclosed unverified-bibliography caveat. The fixture demonstrates the policy catching exactly the kind of case it exists to catch, rather than being set to whatever status looks best.Scope note: "strong novelty language" is gated structurally (
contribution_class == "new_proof"), not via free-text scanning of title/informal_statement prose — that would be a separate NLP-ish follow-up if wanted, not silently covered here.5 failure modes smoke-tested directly, including a positive case (a valid
publication_readyclaim correctly passes every gate). Full corpus (357 packets) + full CI-equivalent pipeline + all issue #6 tooling verified locally, 0 errors.Tracked in
agents/github_issues/DONE.md,FEATURES.md,HANDOFF.md,TRIAGE.md.Roadmap complete: #1-#8 are all closed. 8 commits sit locally on
main, unpushed (this repo was already ~220 commits ahead oforiginbefore this work began) — pushing needs the repo owner's explicit go-ahead given the size.feat(schema): add literature-lineage, idea-attribution, and prior-art…
feat(policy): require contribution, citation, and novelty review for …
feat(corpus): populate proof_profile, dependency_manifest, and litera…