| name | release-readiness-review |
| description | Review freqTLS before a public, CRAN, GitHub, or internal release. |
Release Readiness Review
Use this skill before public releases, GitHub releases, or version tags. Grace
leads; Rose audits claims; Pat and Fisher hold the Definition-of-Done gate.
Required Checks
- Confirm the current branch and a clean working tree.
- Review
NEWS.md, README.Rmd (and the rendered README.md), ROADMAP.md,
docs/dev-log/known-limitations.md, docs/design/46-capability-matrix.md,
and the _pkgdown.yml navigation.
- Run the package checks appropriate to the release:
devtools::document()
devtools::test()
devtools::check()
pkgdown::check_pkgdown() and pkgdown::build_site()
- Check examples and vignettes for unsupported syntax, stale claims, or any
"posterior"/"credible" language describing a freqTLS interval.
- Check dependency versions and platform risks; confirm the normal-Suggests CI
matrix covers Ubuntu R release/devel, Windows R release, and macOS R release,
and remains Stan-free.
- Confirm the benchmark cache provenance is current and the article reads the
cache (no live Stan in CI). Restrict equivalence claims to the matched
relative-threshold, constant-shape model fits; identify the classical
two-stage comparator as an absolute-LT50 approximation.
- Audit every installed data/extdata component against
docs/design/47-data-license-ledger.md, R/data.R, inst/COPYRIGHTS, and
inst/CITATION. Confirm snow-gum remains labelled CC BY-NC 4.0 and that
snow-gum, Kristineberg, or any other permission-pending asset is absent from
the source tarball unless written redistribution authority is recorded.
- Confirm Pieter A. Arnold, Patrice Pottier, and Daniel W. A. Noble have
explicitly agreed to their
aut roles before public release or CRAN upload.
- Build one exact source tarball, record its checksum and inventory, and run
R CMD check --as-cran on that file with normal Suggests. Require zero errors
and warnings and explain every NOTE.
- Before CRAN upload, require the candidate commit's GitHub platform matrix,
win-builder R-devel, R-hub Linux/clang, and independent Grace/Rose/Pat
completion verdicts. Record skipped checks and residual risks in the check
log without describing pending gates as passed.
Release Questions
- What is implemented, documented, and tested?
- What is planned but not implemented (the 0.1.0 non-goals)?
- What should users avoid (weakly identified designs; extrapolated CTmax)?
- What changed since the previous release?
- What would make the release embarrassing if found tomorrow?