What information is needed to rerun a published workflow?

Why

A rerunnable workflow should include a manifest of raw inputs and data access requirements, sample IDs, checksum or content hashes, scripts or workflow definition with commit hash, environment lockfile or container digest, versions of reference genomes or databases and all non-default settings. Document random seeds, hardware assumptions, external service versions and expected intermediate files. Add a minimal example and smoke test.

For instance, an analysis that lists a package version but downloads a changing annotation database each run is not fully pinned. Secure or patient data may require governed access, yet metadata and synthetic tests can still improve transparency. Avoid making FAIR compliance synonymous with perfect reproducibility: discoverability and reusable metadata help, but they do not ensure every dependency executes identically.