Guide · Agent skill
How to use the Scientific Writing agent skill: IMRAD, citations, checklists
Updated · written with AI from the project's own documentation and checked against our catalogue data
The scientific-writing skill is a set of academic writing guidelines that an AI agent follows when you ask it to draft or revise a research paper. It covers IMRAD section structure, paragraph and citation habits, common writing pitfalls, and reporting checklists. This guide explains what the skill does, how to install it, how to trigger it, and where its limits are, so you can use it on a manuscript with confidence.
Scientific Writing Guideaiming-lab/AutoResearchClaw · MIT Download the skill Quick start
- 1Download the ZIP. Get the scientific-writing skill ZIP from the directory page.
- 2Unzip into a skills folder. Extract it to ~/.claude/skills/ for personal use, or to .claude/skills/ inside a project.
- 3Or upload in the Claude apps. Go to Settings, then Capabilities, then Skills, and upload the skill.
- 4Make a request. Ask for help drafting or revising a research paper, naming the section or study type.
- 5Check the reporting guideline. Ask which checklist fits your design, such as CONSORT, STROBE, PRISMA or STARD.
- 6Verify everything. Check every citation and result yourself, then confirm the target journal's author guidelines.
What the scientific-writing skill makes the agent do
An agent skill is a folder containing a SKILL.md file with instructions that the agent reads when a request matches the skill's description. This one is described as academic manuscript writing with IMRAD structure, citation formatting, and reporting guidelines, and it is meant for drafting or revising research papers.
When it applies, the agent organizes the manuscript into Abstract, Introduction, Methods, Results and Discussion. It follows guidance on paragraph structure, citation habits and common pitfalls, and points you to the reporting checklist that fits your study type. The skill is written guidance only: it does not run scripts, use the network, or check or fetch anything.
- Abstract: objective, methods, key results and conclusion in 150-300 words
- Introduction: broad context, then the specific gap, then your contribution
- Methods: enough detail to replicate, in past tense
- Results: findings without interpretation, paired with figures or tables
- Discussion: interpretation, literature comparison, limitations and implications
- Reporting checklists: CONSORT for randomized trials, STROBE for observational studies, PRISMA for systematic reviews, STARD for diagnostic accuracy
Install the scientific-writing skill
Download the ZIP for the skill from the directory page. The skill lives at .claude/skills/scientific-writing in the AutoResearchClaw repository, and the package contains SKILL.md.
For a personal install, unzip the folder into ~/.claude/skills/ so the skill is available across your projects. For a project-only install, unzip it into the .claude/skills/ folder inside that project.
Alternatively, you can upload the skill in the Claude apps. Open Settings, go to Capabilities, then Skills, and upload the skill there.
- Personal: ~/.claude/skills/scientific-writing
- Project only: .claude/skills/scientific-writing inside your project folder
- Claude apps: Settings → Capabilities → Skills
~/.claude/skills/Trigger the skill with natural requests
You do not need a special command. The skill is selected when your request matches its description, which mentions drafting or revising research papers, IMRAD, citations and manuscripts. Plain requests work well, and naming the section or study type helps the agent pick the right checklist.
Give the agent your draft, your target section or your study design. The more specific the request, the more useful the feedback.
- "Help me restructure this manuscript into IMRAD sections."
- "Revise my Methods section so another lab could replicate the study."
- "Check my abstract against the final results and tell me what doesn't match."
- "My study is a randomized trial. Which reporting checklist should I follow?"
- "Run a pre-submission checklist on figure references and citations in this draft."
What the skill needs and how it behaves
The skill lists no required tools or packages. It does not run scripts and does not use the network, so everything it does happens in the conversation with the agent.
Because it only supplies writing conventions, its output is advice and edits to your text. It cannot confirm that a cited source says what you claim, or that your statistics are correct.
- Requires: nothing listed
- Runs scripts: no
- Uses network: no
Tips for getting better results
Paste the full section you want reviewed, plus the abstract if you are checking consistency. The revision checklist in the skill asks for figures and tables to be numbered in order, references to match in-text citations, and the abstract to reflect the final results.
Name your target journal. The skill itself says to check each journal's author guidelines, because word limits, citation style and formatting differ between journals.
- Ask for one section at a time for closer feedback
- Ask for a check of hedging, such as stacked qualifiers in one sentence
- Ask the agent to flag citation strings of five or more references and suggest the most relevant two or three
- Read the revised text aloud to catch run-on sentences
Safety notes and limits
The skill is documentation-style guidance with no scripts and no network access, so there is nothing it runs on your machine or sends elsewhere. Its own content is adapted from K-Dense-AI/claude-scientific-skills, as noted in its metadata.
Treat the output as a style and structure review. Cited sources, results and scientific claims still need human checking against the original papers and your data.
What people use it for
- Restructuring a messy draft into Abstract, Introduction, Methods, Results and Discussion
- Tightening a Methods section so the procedure can be replicated
- Running a pre-submission check on figure numbering, reference matching and abstract accuracy
- Picking the right reporting checklist for a randomized trial, observational study, systematic review or diagnostic accuracy study
Frequently asked questions
How do I install the scientific-writing agent skill?
Download the ZIP from the directory page and unzip it into ~/.claude/skills/ for personal use, or into .claude/skills/ inside a project. You can also upload it in the Claude apps under Settings, then Capabilities, then Skills.
Is the scientific-writing skill free to use?
The source material lists the skill under the MIT licence. Check the licence text in the repository for the exact terms before reuse.
What does the scientific-writing skill require?
The skill lists no required tools. It runs no scripts and uses no network, so it works through the agent conversation alone.
Does the skill check my citations or results?
No. It gives writing and structure guidance and a revision checklist. Cited sources and scientific results still need human verification.
Which reporting checklist should I use?
The skill maps randomized trials to CONSORT, observational studies to STROBE, systematic reviews to PRISMA and diagnostic accuracy studies to STARD. Ask the agent which one fits your design.
Does this skill match my target journal's requirements?
Not fully. The skill tells you to check each journal's author guidelines, since word limits, citation style and formatting vary, so use it alongside those guidelines.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you are a researcher or graduate student drafting or tightening a paper and want consistent IMRAD structure, citation habits and a pre-submission checklist. Skip it if you need the agent to verify scientific claims, or to meet a specific journal's formatting rules, since it gives style advice only.
Download the skill