From a question to research you can share

On this page

Start with a question you care about. Use the steps below as the work develops. You do not need to read the whole hub first.

Try a source comparison without code. The browser guide explains how to submit it or create a project.

1. Find a useful question#

Choose a problem from your field, daily work, reading, or a conversation across fields. Write what you want to understand and who could use the answer.

Ask in Q&A or share an idea. A rough question is enough. Use your preferred language. Sign in to GitHub to post; reading the public guides needs no account. The topic guide and searchable example topics can help. You can also download the topic table as a comma-separated values (CSV) file.

2. Check the closest work#

Find a few closely related papers, reviews, models, or datasets. Read the relevant parts. Record what they established, what remains uncertain, and why your question matters.

Use your experience to question an assumption or connect findings across fields. A different dataset or a low search count alone does not establish a new contribution. An agent can help search and compare sources; check what each source actually supports.

3. Check that the work is possible#

Can you obtain the evidence, data, or model? Do its definitions fit your question? What help, software, time, or computing resources would you need?

Try a small first task before planning a large study. Use the resource guide or ask an administrator for missing access. Narrow or change the question when needed.

4. Create a research repository#

Use the research repository template and name a human lead. An administrator can create a repository directly under the OSRI organization; contributors can also begin in their own account. Name the repository after the paper or working title, using lowercase words separated by hyphens.

Put the full paper or working title at the top of the README, followed by the question, lead, research areas, and current state. Open Issues for tasks and Pull requests for proposed changes. Ask an administrator to add the study to the research directory. One repository can span several areas and produce several publications.

Follow the browser setup steps or ask a maintainer to help. To find collaborators, state the question, first task, and help needed. Record roles in contributions.md and discuss authorship as the work develops.

5. Choose a method and a small set of tools#

Read the closest methods and choose the workflow profile that fits. A source comparison may need only notes and a spreadsheet. Computational work may need one R or Python environment.

You may use your own agent or selected MedSci skills. Check the skill's scope and record its version. The tool catalog is a reference, not an installation list. A manual method can meet the same requirements.

Use protocol.md in the project to record the question, comparison, inputs, methods, and checks. Label exploration as exploration. For a planned test, record the plan before running the test. Explain later changes.

6. Do the work and keep the evidence#

Keep source notes, analysis steps, model settings, and results in the project. Preserve editable files and use open exchange formats.

Check sources, calculations, data transformations, and model behavior as the method requires. Record failed checks and uncertainty. If a first idea fails, record why. Change the question or method openly and retain useful negative results.

7. Ask for review#

Make a small pull request or request review in Q&A. Link the draft or result, evidence, checks, and requested expertise. Ask a maintainer to confirm the reviewer and next action. Keep missing review marked as pending. The browser guide explains pull requests and the assisted route.

Have a person examine the evidence, interpretation, limitations, and credit. Seek the domain and method expertise the claim needs. Report review still needed. An agent's review is assistance; it does not establish independent human review.

8. Write what you learned#

State the question, closest work, method, findings, and limitations. Explain what the contribution changes in our understanding. Link each material claim to its evidence.

Use the project's writing outline. A substantive blog, preprint, full conference paper, or journal article can be a useful output. Choose the audience and venue that fit the completed work. Novelty and acceptance need evidence; they are not guaranteed by this process.

9. Share and record the output#

The human lead authorizes publication of the reviewed result. State its peer-review status and share it in Show and tell.

Follow Publish and record your research to choose a destination, prepare files, submit, address revisions, and preserve the public version. Confirm authorship and contribution credit before submission.

Add its public link and type to the publication register. Link related outputs from the same research work. Keep the evidence and reusable materials available under their applicable terms.

A first task without code#

Choose topic H02: how service simulations represent missed appointments and access. Start with three closely related papers.

Record each paper's question, assumptions, access measure, and supporting passage in a small table. Mark unavailable material as unchecked and request help. Compare what the papers support and what they leave uncertain.

Use the browser contribution route to share the table and ask a maintainer to add it to the research record. No terminal, agent, or software installation is required.

This first comparison is a learning task, not a complete review or a confirmed research gap. If a useful question emerges, define the search and method with collaborators. Then develop the evidence before claiming a novel finding.

View source on GitHub