The Findings Workflow: An agentic, collaborative, provenance-tracked framework for scientific data analysis with Claude Code

The Findings Workflow, a Claude Code plugin, reproduces central results from a plasma extracellular vesicle proteomics dataset for Alzheimer’s disease (Mag-Net) while preserving full provenance.

• Why it matters: It addresses the risk that LLM-assisted analysis can be statistically or computationally wrong and that figure-level conclusions may lack reproducible, inspectable provenance.
• What they did: The plugin uses staged, gated progression with subagents for statistics, programming, visualization, research, reporting, and validation, plus independent review and vetted templates.
• The result: It automatically records findings with data, code, methods, figures, and provenance, enabling trusted inspection and reproducibility of LLM-assisted analyses.

AI summary of the abstract · may contain errors · not medical advice · read the paper.