This checklist defines Permea Core's initial public release checklist.
Use it before public releases of docs, code, schemas, benchmark artifacts, paper-support packages, datasets, preprints, or derivative projects.
Passing this checklist does not itself create legal, clinical, biological validation, or public preprint approval. Approval must be recorded separately where required.
- Release type is identified.
- Release state is identified.
- Files included are listed.
- Files excluded are listed.
- Reviewer roles are assigned where needed.
- Known blockers are documented.
- Public-facing status is accurate.
- Release notes match actual release contents.
- Claims follow
docs/scientific-governance/CLAIM_REGISTRY.md. - Ambition is separated from achieved evidence.
- Benchmark claims state dataset, split, metric, model, artifact, and evidence level.
- Computational evidence is not described as biological validation.
- Wet-lab validation is not claimed unless documented.
- Clinical efficacy is not claimed.
- Universal delivery prediction is not claimed.
- Production-grade drug delivery platform status is not claimed.
- "AlphaFold for Delivery" is used only as ambition or positioning, if used at all.
- Dataset source is documented.
- Source citation or attribution is documented where available.
- Source version or access date is documented where available.
- License or terms are documented where available.
- Redistribution status is documented.
- Derived artifact status is documented.
- Unknown source/license items are marked unresolved.
- No final legal conclusions are made without appropriate review.
Confirm the release does not include restricted or unresolved:
- row-level sequence datasets
- row-level labels
- row-level feature tables
- row-level predictions
- ranking tables
- split manifests
- group assignments
- sequence-pair leakage CSVs
- raw upstream dataset mirrors
- private or partner-controlled records
If any item is included, explicit permission and review must be documented.
- Benchmark contract is identified.
- Dataset reference is identified.
- Split protocol is identified.
- Model or baseline is identified.
- Metrics are identified.
- Result artifacts are listed.
- Run manifest is present or limitation is documented.
- Aggregate and row-level artifacts are separated.
- Exploratory outputs are not presented as contract-grade results.
- Code revision is recorded.
- Config or command is recorded.
- Environment is described.
- Data access constraints are described.
- Reproducibility level is stated.
- Aggregate vs row-level reproducibility is distinguished.
- Rerun limits are documented.
- Reproducibility is not described as biological validation.
- Manuscript title matches claim scope.
- Abstract matches evidence level.
- README claims match manuscript claims.
- Dataset scope and limitations match across materials.
- Data/code availability wording is consistent.
- Supplement does not expose blocked artifacts.
- Source-to-claim review is complete or unresolved items are visible.
- Public preprint status is accurately marked.
- Hold status is not contradicted by any public-facing material.
- README does not overstate paper readiness.
- Website does not overstate evidence level.
- Decks do not imply validation beyond documented evidence.
- Release notes do not broaden benchmark claims.
- Screenshots or figures do not reveal restricted row-level artifacts.
- No secrets, credentials, API keys, tokens, or private keys are included.
- No private data are included.
- No confidential partner data are included.
- No nonpublic source terms are exposed.
- Sensitive findings are reported through private channels.
- Security review is complete where relevant.
- Contributors are acknowledged according to current repo policy.
- Paper authorship is not promised by contribution alone.
- Authorship decisions are deferred to formal policy/manual review where needed.
- Wet-lab partner contributions are reviewed for ownership, confidentiality, and credit.
- Reviewer contributions are recorded where appropriate.
- Maintainer approval recorded.
- Claim review approval recorded.
- Dataset/provenance review approval recorded where needed.
- Reproducibility review approval recorded where needed.
- Paper/source-to-claim approval recorded where needed.
- Founder/manual approval recorded where required.
- Release decision state recorded.
Do not release publicly if:
- public status is still Hold
- source/license/redistribution status is unresolved for included row-level artifacts
- row-level restricted artifacts are included without explicit permission
- secrets or private data are present
- claims exceed evidence level
- dry-lab evidence is described as wet-lab validation
- clinical efficacy is implied
- universal delivery prediction is implied
- AlphaFold-level performance, adoption, or standardization is implied
- README, manuscript, website, supplement, or release notes contradict one another
- required manual approval is missing
Use this template in release review records:
Release type:
Release state:
Files included:
Files excluded:
Claim review:
Dataset/provenance review:
Artifact release review:
Reproducibility review:
Paper/preprint alignment review:
Manual approval:
No-go conditions present:
Final decision:
Known limitations:
Next required action:
Permea may use "AlphaFold for Delivery" as ambition or positioning only.
Permea must not claim AlphaFold-level performance, adoption, or standardization; completed wet-lab validation unless documented; clinical efficacy; universal delivery prediction; production-grade drug delivery platform status; or dataset redistribution permission without source/license approval.
Benchmark claims must remain scoped to dataset, split, metric, model, artifact, and evidence level.
Public release must not include row-level restricted artifacts unless explicit permission and review are documented.