clinical-research-session

M4 Clinical Research Workflow

Structured clinical research from hypothesis through analysis. All work is tracked in a vitrine study. Use the vitrine-api skill when you need full display API signatures or advanced interaction patterns.

When This Skill Activates

• User invokes /research command
• User describes research intent: "I want to study...", "Can we analyze...", "What's the mortality rate for..."
• User mentions cohort analysis, hypothesis testing, or comparative studies

Terminal and Vitrine

The researcher has the terminal and vitrine open side by side. Vitrine is where structured interaction happens — forms, data review, approvals. The terminal is where you discuss, explain reasoning, and refine.

When blocking for input (wait=True), always narrate the handoff in the terminal before the show() call. Tell the researcher what you've posted and what you need: "I've posted the study parameters form in vitrine — please fill in your outcome and exclusion criteria."

Study Setup

Every research session is organized as a study — one study per research question, spanning one or more conversations.

from vitrine import (
    show, section, register_output_dir, study_context,
    list_studies, export, Form, Question,
)

STUDY = "early-vasopressors-sepsis-v1"
output_dir = register_output_dir(study=STUDY)

# FIRST card: study description as the opening vitrine card
show("""# Early Vasopressor Use in Sepsis
...research question, design, key definitions...
""", title="Study Description", study=STUDY)

Continuing a study: Call list_studies() and study_context(study) to re-orient. Use section() to mark a new conversation within an ongoing study — not a new study.

Branching: Create a new version (v2) when the researcher wants a different approach.

Output Structure

Cards tell the story. Scripts ARE the science. Create this structure at study start:

output_dir/
├── PROTOCOL.md
├── RESULTS.md
├── scripts/
│   ├── 01_cohort_definition.py
│   ├── 02_baseline_characteristics.py
│   ├── 03_outcome_analysis.py
│   └── ...
├── data/
│   ├── cohort.parquet
│   ├── baseline_table.parquet
│   └── ...
└── plots/
    ├── age_distribution.json
    ├── kaplan_meier.json
    └── ...