Step 2. Run Your First Pipeline

Run a full end-to-end research pipeline using sklearn's built-in breast cancer dataset.
No data download required — you can start right away.

What this demo produces

Manuscript draft

IMRAD structure, ~2,200 words, DOCX/PDF

4 Figures

ROC curve, Confusion Matrix, 300 DPI

STARD audit report

30-item checklist with revision suggestions

Presentation slides

12-slide PPTX with speaker notes

1

Enter the prompt in Claude Code Desktop

Open Claude Code Desktop and type the following:

/orchestrate Run an end-to-end diagnostic accuracy study using sklearn's load_breast_cancer data. Include analysis, figures, manuscript draft, STARD audit, and presentation slides.

You can also type in Korean — Claude understands both languages.

2

Watch the pipeline run automatically

Claude chains the following skills in sequence. When a permission prompt appears, click "Allow" to continue.

  1. 1.analyze-stats — loads the data, generates Table 1, trains and compares 3 models (LR, RF, SVM)
  2. 2.make-figures — produces 4 figures: ROC curve, Confusion Matrix, Calibration Plot, and more
  3. 3.write-paper — drafts a full IMRAD manuscript (Title, Abstract, Introduction, Methods, Results, Discussion)
  4. 4.check-reporting — audits the draft against the STARD 2015 checklist (30 items)
  5. 5.present-paper — generates a 12-slide PPTX deck with speaker notes

Total time: approximately 5 minutes (varies with network speed)

3

Review the output

Once the pipeline finishes, Claude will show you where the files are saved — typically in your current working folder.

Key files:

  • manuscript_draft.docx — manuscript draft
  • figures/ — ROC, Confusion Matrix, and other 300 DPI images
  • stard_compliance_report.md — STARD audit results
  • presentation.pptx — presentation slides
4

Try it with your own data

If the demo worked well, try running a pipeline on your own data:

Have a CSV or Excel file?

/orchestrate Run a diagnostic accuracy study using data.csv

Have a manuscript draft?

/check-reporting Audit manuscript.docx against the STROBE guideline

Only have a topic so far?

/search-lit Search PubMed for literature on "AI lung nodule diagnostic accuracy"
← Previous: InstallationNext: Skills by Scenario →