dna-seq/manuscript-skills

Skills that help scientists prepare manuscripts for publication, for
Claude Code and Codex (AI assistants
that work with the files on your computer). They review a draft, check its references and figures,
simulate the referees of a target journal, edit the prose, choose and prepare for a journal, and answer
the referees after a decision. They work with manuscripts written with or without AI assistance.
Start with manuscript-improve. One request runs the relevant checks and edits across the whole paper
and ends with a corrected version and a dashboard of what the run found. For a single task, use
one of the sixteen focused skills.
The aim is to help you publish sound research with less avoidable delay, expense and paperwork. You
decide which journal to pursue and how to handle uncertain or burdensome requirements. Made by
DNA Seq and released under the MIT licence.
## What a run gives you
Every manuscript-improve run ends with manuscript-dashboard.html, one file that opens offline in a
browser. The dashboard above is from a run on draft v0.3 of a real software paper, Just-DNA-Lite. The
run found 114 confirmed problems, each counted once however many skills found it, and sorted them by
what to fix now. A co-author had reviewed the same draft separately; the run saw neither that review nor
any later version of the paper. Of the co-author's 125 points, the run found 79 (27 of them only in
part) and missed 46.
The dashboard is self-contained, so you can send it to a colleague as it is. The same review also comes
as a Word document: the corrected version as tracked changes signed by each stage, with the problems,
the questions for you and the reason for each edit as margin comments. Google Docs opens it with the
changes as suggestions. No question for you is ever written into the paper's text, where it could be
submitted by mistake.
The paper itself , each problem highlighted where a skill quoted it, with what is wrong and what to do beside it.
The corrected version as track changes: routine and evidence-backed fixes applied, the other changes proposed, all of it yours to accept in one step or one by one.
The journals chosen by your priorities, each AI rule shown as an icon, and a red flag for rules that can cost you, such as a ban on AI drafting or an AI detector.
The simulated reviewers : a recommendation per journal, and every concern with its suggestion, whether it held up when checked against the paper, and who can act on it.
## How it differs from other academic-writing skills
Several popular skill collections also review, edit and reference-check papers. Compared with the
most-starred of them in October 2026 (the comparison),
these parts have no counterpart there:
Which rules apply to the study. manuscript-policy-fit
works out which clinical, ethics and privacy requirements apply, including to authors' own public
genomes and to papers referred between bioRxiv and medRxiv, and drafts the clarification.
Journals ordered by the authors' priorities, AI rules included.
manuscript-journal-select ranks preprint servers and
journals by speed, reach, cost and how strict their AI policies are, with the deciding rule quoted and