AGENT SKILL

r-publish

Orchestrate the full R package publication pipeline: CRAN (and R-universe) for distribution, JOSS for a peer-reviewed DOI, optionally a JSS methods paper. Use for "publish my R package". For a Python package use pypi-publish; for a standalone LaTeX paper (no DESCRIPTION) use the papermill plugin instead.

License 许可
MIT
最近更新
2026-10-23
Stars
★ 0
Forks
0

01 前言 Front Matter

name
r-publish
description
Orchestrate the full R package publication pipeline: CRAN (and R-universe) for distribution, JOSS for a peer-reviewed DOI, optionally a JSS methods paper. Use for "publish my R package". For a Python package use pypi-publish; for a standalone LaTeX paper (no DESCRIPTION) use the papermill plugin instead.
argument-hint
[package path]

03 SKILL.md 正文

R Package Publication Pipeline

Orchestrate CRAN submission, then a JOSS paper, then optionally a JSS methods paper. This skill coordinates /pub-pipeline:cran-audit, /pub-pipeline:joss-audit, and /pub-pipeline:joss-draft into one workflow. Package root: $ARGUMENTS if given, else the current directory. If it has no DESCRIPTION with a Package: field, say so: point to /pub-pipeline:pypi-publish for a pyproject.toml project, or the papermill plugin for a .tex paper.

Strategy

  1. CRAN: distribution, discoverability, install.packages().
  2. JOSS: quick DOI, peer-reviewed citation, Crossref/Google Scholar indexing.
  3. JSS (optional): high-prestige methods paper, Web of Science indexed.

Each stage builds on the previous: CRAN gives legitimacy, JOSS a citable DOI, JSS academic weight for the methodology.

Workflow

Phase 1: Assess Current State

Load user config (Read tool): read .claude/pub-pipeline.local.md if it exists. Extract author metadata, r.domain, r.audience, r.competitors, r.targets, r.universe, r.dependency_chain, and related_work. Companion papers and preprints in related_work shape the JOSS paper (cite them) and the overall strategy. If the file is missing, offer to create it from ${CLAUDE_PLUGIN_ROOT}/docs/user-config-template.md.

  1. Read DESCRIPTION for package metadata (Read tool).
  2. Check git history (Bash tool).
  3. Look for existing paper.md, CITATION.cff (Glob tool).
  4. CRAN status: fetch https://cran.r-project.org/package={name} (WebFetch tool).
  5. R-universe status (Bash tool, structured JSON; {user} is r.universe, skip if unset): curl -s https://{user}.r-universe.dev/api/packages/{name} returns Version, _registered, and _status (build result); a 404 means the package is not in that universe.
  6. JOSS status: search joss.theoj.org (WebSearch tool).

Dependency chain (Bash tool): for each non-base, non-recommended package in Imports, Depends, and LinkingTo:

Rscript -e 'ap <- available.packages(); cat("PKG_NAME" %in% rownames(ap))'

A dependency missing from CRAN is a hard blocker for R CMD check --as-cran. Present the submission order (dependencies first), cross-check r.dependency_chain, and offer to switch to the dependency package.

CI status: if .github/workflows/ exists (Glob tool), run gh run list --limit 3 --json conclusion,name,headBranch (Bash tool). Failing CI undermines both CRAN confidence and JOSS reviewer trust.

Classify the package:

State Next Step
Dependencies not on CRAN Start with the dependency package
On R-universe only, not on CRAN CRAN audit; R-universe already gives installable binaries, so JOSS can proceed in parallel
Not on CRAN, no paper.md Start with CRAN audit
On CRAN, no paper.md Start with JOSS audit
On CRAN, has paper.md Review paper, prepare JOSS submission
On CRAN, published in JOSS Consider JSS (if substantive methodology)

Bioinformatics packages (genomics, proteomics) usually belong on Bioconductor instead of CRAN, which has its own submission process.

Present the assessment and confirm the plan before proceeding.

Phase 2: CRAN Readiness

Run /pub-pipeline:cran-audit. If the package is not ready: present the gap report, fix automatable issues, guide manual fixes, re-audit. Skip to Phase 3 if already on CRAN.

Submission steps (when ready):

  1. usethis::use_version("minor") to bump the version (commit pending changes first; it prompts otherwise).
  2. devtools::check(args = "--as-cran") for the final check.
  3. Other platforms: devtools::check_win_devel() (win-builder) and rhub::rhub_check() (R-hub v2; one-time rhub::rhub_setup(); rhub::rc_submit() for non-GitHub repos).
  4. Submit (user runs this): devtools::submit_cran() asks interactive confirmation questions that the Bash tool cannot answer. Hand it to the user to type in the prompt:
    ! Rscript -e 'devtools::submit_cran()'
    
    or to run in an interactive R session. Then they confirm via the emailed link.
  5. After acceptance: usethis::use_github_release(), then usethis::use_dev_version().

If CRAN rejects: read the email carefully, fix every identified issue (not a subset), re-run devtools::check(args = "--as-cran"), note the fixes in cran-comments.md, and resubmit. Do not resubmit more than once a week unless CRAN asks.

Phase 3: JOSS Readiness

Run /pub-pipeline:joss-audit. Common gaps: under 6 months of public history, no CONTRIBUTING.md, no paper.md. If there is no paper, go to Phase 4.

Phase 4: Draft JOSS Paper

Run /pub-pipeline:joss-draft. Then present the draft, iterate on feedback, re-run /pub-pipeline:joss-audit for a full checklist review, and commit the paper to the repository.

Phase 5: JOSS Submission

  1. Pre-submission: package installable (ideally on CRAN), paper.md and paper.bib committed, audit passes, ORCID for all authors.
  2. Submit at https://joss.theoj.org/papers/new with the repository URL; wait for the pre-review issue.
  3. During review (typically 4-12 weeks): respond promptly, make requested changes, re-run the audit.
  4. Post-acceptance: tag a release; archive it on Zenodo (enable the Zenodo GitHub integration for the repository, publish a GitHub release with gh release create, copy the deposit DOI); post the version and archive DOI in the review issue. JOSS then assigns the paper DOI via Crossref.

Phase 6: JSS Consideration (Optional)

Appropriate when the package implements novel statistical methodology with substantive mathematical content, the paper would be 20-50 pages of methods plus software description, and the audience is academic statisticians.

Not appropriate when the package is primarily a software tool (JOSS covers it), the mathematics is textbook material, or the paper would just be a longer JOSS paper.

If appropriate: JSS uses the rticles::jss_article format, is indexed in Web of Science, and reviews take 6-18 months. The papermill:writer agent can draft long-form academic papers.

Decision Helpers

Criterion Minimum for CRAN Minimum for JOSS
License OSI-approved OSI-approved
Tests Pass R CMD check Automated test suite
Docs Man pages for exports README, examples, API docs
History Any 6+ months public, sustained
Users Not required Research impact evidence
Paper Not required 750-1750 word paper.md
Stage Duration
CRAN prep and audit 1-3 days
CRAN review 1-7 days (usually 1-2)
JOSS prep and paper 1-2 days
JOSS review 4-12 weeks
JSS prep and paper 2-4 weeks
JSS review 6-18 months

Reference Files

  • ${CLAUDE_PLUGIN_ROOT}/docs/cran-reference.md: CRAN Repository Policy, submission workflow, common rejections
  • ${CLAUDE_PLUGIN_ROOT}/docs/joss-reference.md: JOSS requirements, reviewer checklist, paper format

Important Notes

  • Never submit to JOSS before the package is installable (ideally on CRAN).
  • CRAN and JOSS are complementary: CRAN for distribution, JOSS for citation.
  • The user controls the pace: present options, don't push.

04 同仓库更多 Skills

Abdulqudduz
Abdulqudduz/claude-anvil

worldsmith-methodology

Use when working in a project with a worldbuilding documentation ecosystem (docs/ or lore/ directory). Provides the documentation-first methodology: canonical hierarchy, doc roles (timeline, lore, characters, systems), propagation discipline, and the dual canonical and exploratory workflow.

★ 0
Abdulqudduz
Abdulqudduz/claude-anvil

prose-craft

Use when writing fiction, editing fiction prose, reviewing chapters, or drafting scenes. Covers prose craft, dialogue mechanics, scene structure, and common failure modes in AI-assisted fiction. Trigger phrases: "write fiction", "edit chapter", "prose review", "scene draft", "dialogue check", "fiction craft", "edit this scene", "review my prose".

★ 0
Abdulqudduz
Abdulqudduz/claude-anvil

vista

This skill should be used when the user asks about research directions, open problems, future work, or follow-up research from the academic literature. Trigger phrases include "open problems in X", "what's next for Y", "future work for paper Z", "research directions on T", "salient follow-up research", "what should I work on next in", "find me follow-up research unrelated to my prior work", "broad survey of W", "neglected directions in V". Routes the request to the right Vista MCP tool, reads back the structured paper sections, and synthesizes research directions in the conversation.

★ 0
Abdulqudduz
Abdulqudduz/claude-anvil

synthesize

Make a research-agent run conclude and write synthesis.md (outcome, key findings, failed approaches, open questions, recommendations). Signals a live researcher to wrap up, or launches one in synthesis mode when none is running. Use when results are good enough or the agent is stalling.

★ 0
Abdulqudduz
Abdulqudduz/claude-anvil

status

Show the current state of a research-agent run from .research/: liveness, cycles and budget, sub-problem and hypothesis statuses, eval trend, current focus, and branches. Read-only.

★ 0
Abdulqudduz
Abdulqudduz/claude-anvil

resume

Resume an interrupted research-agent run from .research/. Re-launches the researcher to reload its structured state and continue from the current focus, optionally on a named branch or with a new budget. Use after a session restart, crash, or explicit pause.

★ 0