Quick Start

You do not learn Researcher by reading a manual. You learn it by talking to Claude the way you would talk to a very organized, slightly caffeinated collaborator. Here is a first session.

1. Start a paper

I want to write a paper on self-supervised learning for ECG arrhythmia classification.

The plugin scaffolds a manuscript/ folder (per-section .tex files, config.yaml, a bibliography, figures and tables folders), asks a couple of clarifying questions, and you are ready.

2. Research the topic

Do a systematic literature search on self-supervised learning for ECG classification.
/researcher:fact-check Self-supervised pretraining can match supervised ECG performance within one percent.
/researcher:sota PTB-XL arrhythmia classification

Every result carries a real, resolvable citation. The fact-check comes back with a verdict (Supported, Contested, Partially Supported, Contradicted, or Unsupported) and the evidence behind it.

3. Design and draft

/researcher:design-experiment Does contrastive pretraining improve label efficiency on PTB-XL?
/researcher:draft-section introduction
Create a diagram of my two-stage pretraining and fine-tuning pipeline.
Turn this results CSV into a booktabs table with the best result in bold.

4. Review and submit

/researcher:review-paper
/researcher:find-journal --filters Q1, open access
/researcher:submit-ready

The review runs a five-persona panel with rubric scores. /researcher:submit-ready runs a reporting checklist (citations resolve, formatting, word counts, required sections) and reports pass or fail with fix instructions. With the optional evidence kernel installed, it is also gated on a passing evidence-lineage compile (shipped in 0.4.0): a refusal-grade defect, such as an orphan claim or an altered number, blocks the ready verdict until it is fixed.

5. Handle revisions

/researcher:revise R1
Here are the reviewer comments: [paste]
Generate the response to reviewers document.

That is the whole loop. For worked, end-to-end recipes with real output, open the Cookbook.