data-availability

Data Availability

Overview

Use this skill to turn manuscript-supporting data into a submission-ready availability package: statement text, repository plan, source-data mapping, and unresolved-risk flags.

This skill is narrower than submission-audit and more operational than scientific-writing. Use it when the bottleneck is no longer the paper's story, but whether the data-sharing package is specific, durable, and journal-ready.

When To Use

Use this skill when:

• the manuscript needs a Data Availability statement
• accession numbers, DOIs, repository records, or source-data files are missing or unstable
• the paper mixes generated data, reused public data, and restricted data
• "available upon request" wording needs to be tightened or replaced
• a submission or revision needs data-sharing text that editors can audit quickly

Do not use this skill for:

• generic section rewriting unrelated to data sharing
• full manuscript restructuring
• bibliography cleanup that does not affect dataset availability

Workflow

1. Identify the target journal and article type.
2. Inventory the datasets that support the main and supplementary claims:
   • raw data
   • processed data
   • figure source data
   • reused public datasets
   • code-adjacent outputs that are necessary to inspect the results
3. Assign each dataset one access route only:
   • public repository
   • controlled-access repository
   • within paper or supplement
   • reused public source
   • third-party restricted
   • justified request route
4. Choose repository and identifier strategy before drafting the statement.
5. Draft explicit dataset-to-location wording.
6. Check that the statement covers both newly generated and reused data.
7. Flag unresolved fields rather than inventing repository names, accession IDs, DOIs, or access rules.

Working Rules

• Prefer public, discipline-specific repositories when possible.
• Treat available upon request as weak unless a real legal, ethical, commercial, or third-party restriction exists.
• If data cannot be public, state:
  • why access is restricted
  • who controls access
  • how requests are reviewed
  • what metadata, source data, or derived data remain public
• Do not merge data, code, protocols, and materials into one vague availability sentence unless the journal explicitly wants that.
• Do not invent accession numbers, DOIs, repository names, reviewer links, or embargo terms.
• List only the resources the analyses actually used, and say plainly where they came from ("All response data were obtained from [collection], [DOI]"). Resources the study did not use are not listed, and an identifier the source does not record is left out rather than confessed as "not exposed" or "could not be identified"; flag such gaps to the author instead.
• Promise nothing: state the routes and files that exist now. "On request" and "Source Data files" appear only when that route or file is real, and a future release or arranged access ("will be released", "planned before acceptance", "can be arranged") is not stated at all.
• Name the scientific content, not the repository layout: no folder paths (source_data/), file mechanics or workbook instructions in the statement.
• Name the public repositories actually used, for example a Hugging Face dataset for processed inputs and a GitHub repository for code, and keep the Code availability statement consistent with what the released repository contains.
• A Supplementary Data table of sources has one row per resource used, no column that is identical in every row, and publications taken from the manuscript's own bibliography so the two cannot disagree. Generate it from the analysis tables where possible.