domain_id	title	state	short_reason	evidence	limitation	permitted_use	prohibited_inference	next_gate
REL-01	Software & release integrity	HIGH_ENGINEERING_CONFIDENCE	Build, route, schema and cross-surface gates pass locally.	full vinext production build PASS | 51 rendered contract tests | 246 multidisease checks | 24 cross-surface checks | checksum-addressed release objects	Anonymous command-line production QA is blocked by the existing edge HTTP 403 policy.	Rebuild, audit and navigate the released research resource.	Engineering integrity does not validate a biological prediction.	Add authenticated post-deploy route checks to the immutable production receipt.
REL-02	Identity, provenance & data contracts	HIGH_RESOURCE_CONFIDENCE	Identifiers, sources, contexts, nulls and transformations are explicit.	17,921 released gene records | HGNC alias map | Dataset Cards | typed API | checksums and release identities	A valid data object can still be unsuitable for a particular prediction task.	Identity resolution, evidence lookup and reproducible data intake.	Presence in the resource is not disease relevance or efficacy.	Require ontology and inferential-unit validation on every newly imported dataset.
REL-03	Disease evidence synthesis	BOUNDED_RESEARCH_CONFIDENCE	Source-linked context is strong enough for hypothesis and study design.	DMD measured-evidence release | four disease modules | cell-context map | process and evidence graphs	Observational and correction context does not establish candidate-conditioned causal response.	Mechanism navigation, gap analysis and prospective study design.	Do not treat integrated context as target efficacy or causal proof.	Independent replication plus matched perturbation and functional endpoints.
REL-04	Same-context perturbation benchmark	LIMITED_TASK_CONFIDENCE	Ridge slightly improves mean RMSE, but response direction is unsupported.	MRUN-RIDGE-SAFE-2.3-G0-REPEATED-FOLD | RMSE 0.1175 vs train mean 0.1180 | strict outcome gate 0/16	The output is a 2,000-feature aggregate HepG2 mean vector, not a cell population or DMD response.	Permanent same-assay comparator and leakage diagnostic.	No directional transfer, DMD prediction or functional effect.	Beat matched means on perturbation-specific delta and distribution metrics across prospective holdouts.
REL-05	Cross-context & DMD response prediction	NOT_ESTABLISHED	No model has passed a matched DMD candidate-response task.	GEARS failed five frozen splits | scGPT lost to same-coverage baselines | TxPert calibration pending | direct DMD candidate outcome 0/21	The required donor- and cell-state-matched perturbation outcome matrix does not exist in the release.	Method research, task design and governed abstention.	No candidate ranking, efficacy prediction or virtual DMD response population.	Acquire and freeze matched DMD perturbation outcomes before model selection.
REL-06	Therapeutic, patient & clinical use	PROHIBITED	No clinical calibration, prospective outcome set or authorization exists.	zero calibrated DMD models | zero prospective predictions | zero measured outcomes | DOI and institutional licences pending	The platform is a research resource, not a medical device or treatment recommender.	None beyond research hypothesis and study-design support.	No diagnosis, prognosis, treatment selection, dosing or patient-level digital twin.	A separate clinically governed development and validation program; current research gates cannot unlock it.