Alterlab preprint deposition
Skill AlterLab-IEU/AlterLab-Academic-Skills/skills/faculty-life/alterlab-preprint-deposition
Drives preprint deposition across servers (arXiv, bioRxiv, medRxiv, SSRN, OSF Preprints): picks the right server by field, prepares submission metadata, sets the license (arXiv offers CC BY/BY-SA/BY-NC-SA/BY-NC-ND 4.0, the arXiv non-exclusive license, or CC0; bioRxiv/medRxiv offer CC BY/BY-NC/BY-ND/BY-NC-ND/CC0 or No-reuse), maps arXiv category taxonomy, handles immutable versioning and preprint DOIs, checks a journal's preprint/self-archiving policy via the Sherpa Romeo v2 API, and links the posted preprint to the published article. Reuses alterlab-arxiv and alterlab-biorxiv for metadata and alterlab-open-science for data-repository choice. Use when depositing a preprint, choosing a preprint server, preparing an arXiv or bioRxiv submission, setting a preprint license, or checking journal preprint policy; for Zenodo/Dryad/Figshare data deposition prefer alterlab-open-science, for TÜBİTAK Aperta prefer alterlab-aperta. Part of the AlterLab Academic Skills suite.From its SKILL.md
npx -y skills add AlterLab-IEU/AlterLab-Academic-Skills --skill alterlab-preprint-depositionAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
What its file declares
Copied from the file, not written here
The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
12.4 KB, ~2.8k tokens by cl100k_base, as published. Nobody here has run it
Preprint Deposition — Pick a Server, Prepare the Submission, Link the Published Version
The write/deposit side of preprinting. Given a finished manuscript, this skill turns "I want to post a preprint" into a concrete, server-specific plan: which server fits the field, what metadata and category to enter, which license to choose (and what it commits you to), how versioning and the preprint DOI work, whether the target journal even allows a preprint, and how to link the preprint to the article once it is published.
It is deliberately the deposit-side complement to the read-side connectors
alterlab-arxiv and alterlab-biorxiv (which search and fetch preprints) and
to alterlab-open-science (which chooses data repositories and writes DMPs).
This skill does not reimplement their search/metadata code — it calls them.
Quick Start
I finished a CS paper — how do I post it to arXiv? Which category and license?
Should this biology manuscript go on bioRxiv or medRxiv?
Does Elsevier's journal X allow me to post a preprint before submission?
My preprint just got accepted — how do I link the published DOI to the arXiv version?
What license should I pick on arXiv if I might publish in a closed journal later?
→ Identify the field → route to a server (see the matrix) → assemble metadata → choose a license → run the journal-policy check before posting → after acceptance, link the published DOI back to the preprint.
When to Use This Skill
Use this skill when the user wants to act on a preprint — post it, choose where, prepare its metadata/license, or reconcile it with a journal or a published version. Core jobs:
- Server selection — match field + manuscript type to arXiv, bioRxiv,
medRxiv, SSRN, or OSF Preprints. See
references/server_selection.md. - Submission metadata — title, authors + ORCID, abstract, **arXiv primary
- cross-list categories** or bioRxiv/medRxiv subject collection, funding,
declarations. See
references/submission_metadata.md.
- cross-list categories** or bioRxiv/medRxiv subject collection, funding,
declarations. See
- License choice — pick from each server's actual license set and explain
the downstream commitment (e.g. CC BY is irrevocable; a later journal may
object to a permissive preprint license). See
references/licensing.md. - Versioning & DOI — arXiv versions (v1, v2, …) are immutable and permanent; a withdrawal is a new version with a tombstone, never a deletion; bioRxiv/medRxiv assign a DOI on posting and accept revisions.
- Journal preprint policy — query the Sherpa Romeo v2 API for the
target journal's prearchiving (preprint) policy before posting. See
references/journal_policy.md. - Post-publication linking — connect the preprint to the published article
(publisher field on the server; the keyless
api.biorxiv.org/pubs/endpoint surfaces bioRxiv→journal links).
Does NOT Trigger — route adjacent requests to the right sibling
| The request is really about… | Route to | Why not here |
|---|---|---|
| Searching / fetching existing arXiv preprints, resolving an arXiv ID | alterlab-arxiv | Read-side connector; this skill deposits, it does not search |
| Searching / fetching existing bioRxiv preprints for a lit review | alterlab-biorxiv | Read-side connector |
| Choosing a data repository (Zenodo, Dryad, Figshare), writing a grant DMP, preregistration, FAIR | alterlab-open-science | That is data/DMP/repository policy, not manuscript preprinting |
| Depositing to TÜBİTAK Aperta, the açık bilim mandate, a VYP | alterlab-aperta | National Turkish open-science track with its own embargo rules |
| Whether cited references actually exist / hallucinated DOIs | alterlab-citation-verifier | Existence-verification, not deposition |
| Dead-link / 404 checks across a bibliography | alterlab-link-health | HTTP reachability, not preprint posting |
| Picking a target journal / formatting for journal submission | alterlab-open-science (OA route) then the journal's own guide | This skill only checks whether a journal permits a preprint |
This skill answers "how and where do I deposit this manuscript as a preprint, and is that compatible with my journal plans?" It makes no claim about manuscript quality, novelty, or whether the work should be published.
Server Matrix (summary — full detail in references/server_selection.md)
| Server | Field fit | DOI on post | Default-ish license note | Moderation |
|---|---|---|---|---|
| arXiv | physics, math, CS, quant-bio (q-bio), q-fin, stat, EE/sys (eess), econ | No native DOI (arXiv ID is canonical; DataCite DOIs available) | arXiv non-exclusive license, or CC BY / BY-SA / BY-NC-SA / BY-NC-ND 4.0 / CC0 | Moderation + endorsement for new submitters |
| bioRxiv | life sciences / biology | Yes (CSHL-issued DOI) | CC BY / BY-NC / BY-ND / BY-NC-ND / CC0 / No-reuse | Basic screening |
| medRxiv | clinical / health sciences | Yes (CSHL-issued DOI) | Same license set as bioRxiv | Screening incl. ethics/▲non-trial checks |
| SSRN | social sciences, economics, law, humanities | DOI varies by network | Author selects; SSRN posting terms | Light screening |
| OSF Preprints | multi/cross-disciplinary + community servers | Yes (DOI via OSF) | CC0 / CC BY / CC BY-NC-ND / No license | Per-provider |
Category/license rows above name the real option sets each server presents at submission. License implications (irrevocability, journal friction) are in
references/licensing.md; never assert a "best" license without stating the trade-off.
Workflow
1. Determine the field and route to a server
Read the manuscript's domain. STEM-formal (physics/math/CS/stat/eess/econ/q-bio/
q-fin) → arXiv. Biology → bioRxiv. Clinical/health → medRxiv.
Social science/law/economics → SSRN (or arXiv econ). Cross-disciplinary or a
field-specific community server → OSF Preprints. Edge cases and the full
decision tree live in references/server_selection.md.
2. Run the journal-policy check FIRST (if a target journal is known)
Before posting, confirm the intended journal permits preprints. Use
scripts/journal_policy.py to query the Sherpa Romeo v2 API
(https://v2.sherpa.ac.uk/cgi/retrieve, item-type=publication, requires a
free api-key). Report the journal's prearchiving (preprint) permission,
any conditions (embargo, version allowed, required statement), and link the
source. If no key is available, fall back to WebFetch on the publisher's policy
page and say so. Details: references/journal_policy.md.
Most major publishers permit preprints, but conditions vary (some bar posting the accepted version, some require a DOI link or a specific notice). Never assert a policy from memory — verify it per journal.
3. Assemble submission metadata
Build the metadata block the server needs: title, all authors with ORCID and
affiliations, abstract, arXiv primary category + optional cross-lists (or
bioRxiv/medRxiv subject collection), declarations (competing interests, funding,
data/code availability, ethics/IRB for medRxiv), and the manuscript PDF. The
canonical field-by-field checklist is in references/submission_metadata.md.
To look up an existing arXiv/bioRxiv record's metadata for reuse, defer to
alterlab-arxiv / alterlab-biorxiv rather than re-querying here.
4. Choose the license deliberately
Present the actual license set for the chosen server (see the matrix), then explain the commitment:
- CC BY 4.0 — maximum reuse; irrevocable; some closed-access journals dislike a permissive preprint and may ask you to change it (you cannot revoke CC BY on already-posted versions).
- arXiv non-exclusive license 1.0 — arXiv-specific; you keep copyright, grant arXiv a distribution license; the least journal-friction option on arXiv.
- CC BY-NC-*/-ND — narrower reuse; check it against any funder open-access mandate (e.g. cOAlition S Plan S generally requires CC BY).
- CC0 — public-domain dedication; broadest, also irrevocable.
Confirm whether a funder mandate forces a specific license before recommending.
Full table + funder-mandate notes: references/licensing.md.
5. Post, then manage versions
After posting: record the arXiv ID / preprint DOI and the version. Revisions are replacements (arXiv) or new versions (bioRxiv/medRxiv) — the old version stays public and immutable. Do not advise "deleting" an announced arXiv paper; that is impossible — only a withdrawal-version with a tombstone.
6. Link the published article after acceptance
When the paper is published, link the DOI back to the preprint (the server's
"published in" / publisher field; bioRxiv/medRxiv auto-detect many links and
expose them via api.biorxiv.org /pubs/{server}/...). This makes the version
of record discoverable from the preprint and vice versa.
Scripts
scripts/journal_policy.py— query the Sherpa Romeo v2 API for a journal's preprint/self-archiving policy by ISSN or title (needs a free--api-key; prints a structured summary; degrades to a manual-check instruction offline).scripts/server_recommender.py— given a few flags (field, has-clinical-data, target-journal-known, needs-DOI), prints a recommended server + license shortlist with the trade-offs, from the rules inreferences/server_selection.md.scripts/preprint_link_check.py— query the keylessapi.biorxiv.org/details/and/pubs/endpoints to confirm a bioRxiv/medRxiv DOI exists and surface any detected preprint→published-article link.
All scripts are stdlib-first (use requests if present, else urllib), run in a
bare uv run python, and never require a key except the Sherpa Romeo lookup.
Self-Check Before Reporting
- Did I verify the journal's preprint policy (Sherpa Romeo or the live publisher page), or did I assert it from memory? Only the former is allowed.
- Did I name the server's real license options and state the irrevocability / journal-friction trade-off, not a bare "use CC BY"?
- Did I route a search/fetch request to
alterlab-arxiv/alterlab-biorxiv, and a data-repository/DMP request toalterlab-open-science, and a TÜBİTAK request toalterlab-aperta, instead of handling it here? - Did I avoid telling the user to "delete" an already-announced arXiv preprint?
References
references/server_selection.md— full server decision tree, field-by-field.references/submission_metadata.md— per-server metadata field checklist.references/licensing.md— license option sets, irrevocability, funder mandates.references/journal_policy.md— Sherpa Romeo v2 API usage and policy reading.
Part of the AlterLab Academic Skills suite.
What ships with it: 8 files
37.4 KB alongside SKILL.md, 3 of them executable
evals/
- evals.json6.6 KB
references/
- journal_policy.md2.8 KB
- licensing.md3.5 KB
- server_selection.md4.5 KB
- submission_metadata.md3.0 KB
scripts/
- journal_policy.pyruns6.1 KB
- preprint_link_check.pyruns4.8 KB
- server_recommender.pyruns6.2 KB