Start here
Project overviewFirst successful revisionTask recipesInstall & maintain
Agent-assisted installationCodex setupClaude Code setupUpdate, back up and restoreUnderstand the workflow
The complete workflowSection contractsThe method-paper sampleAll skillsFigures & evidence
Figure production and checksWriting principlesBehavior evaluation scopeCompatibility & boundaries
Environments and file formatsChatGPT Work availabilityHelper-script boundariesContribute & learn more
ContributingChangelogAttributionLicensingA few useful answers.
Is this a website that writes my paper?
No. This site introduces reusable instructions and helper scripts for your agent. It does not upload your manuscript or call a model. Actual processing happens in the agent environment you choose.
Do I need to learn every skill name?
Start with paper-workflow or copy a task prompt. The workflow chooses relevant steps. A small paragraph edit does not require the entire manuscript pipeline.
Does installing the skills install all the tools?
No. Figure rendering, source retrieval, DOCX editing and LaTeX compilation require suitable tools in your agent environment. Skills do not supply missing experiments or dependencies.
Can I use it in ChatGPT Work on the web?
Persistent account registration and new-chat invocation remain unverified for this repository. A supported workspace registration or distribution flow is required; files in a task sandbox alone are not an account installation.
What do the tests demonstrate?
Repository tests cover scripts, installation and content consistency. The synthetic teaching example is a reference revision, not a live run or comparative writing-quality benchmark. Behavior observations have their own documented limits.
Is this affiliated with Nature Portfolio?
No. It is an independent open-source project, not affiliated with or endorsed by Nature Portfolio or Springer Nature. Journal-specific instructions take precedence, and no acceptance outcome is predicted.