research-data-availability

Register each dataset with its version, path, access level, license, and hash, then mark sharing status as public, private, restricted, or TBD. The skill drafts the final availability statement from these records and documents why data cannot be shared when constraints apply.

Map each CLM-* claim to its source data, processed data, generating script or notebook, and output artifact. Datasets are registered as DATA-* records and connected to claims, experiments, and figures through the evidence promotion policy.

P0 failures occur when a core claim has no source data or output artifact, or when a data path is fake or contradicted by experiment records. P1 issues include missing hashes, access levels, licenses, or non-reproducible processed data.

Yes. Datasets can be marked private or restricted with the reason recorded, covering privacy, license, advisor, or project constraints. Private datasets and credentials are never stored in git.

No. Data availability is an audit gate, not a substitute for reproducibility. Missing experiment details are routed to the research-experiment-engineering skill, and unresolved final issues go to research-final-audit.