{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/37b670a9280a4a57ae2ac06abb16f8ac\" frameborder=\"0\" width=\"1280\" height=\"960\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":960,"width":1280,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":960,"thumbnail_width":1280,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/37b670a9280a4a57ae2ac06abb16f8ac-30592d72741debc0.gif","duration":289.695,"title":"Building Reproducible Agent Workflows for Research","description":"This Loom explains an agent driven workflow for documenting and generating research figures from code and data. The author describes a canvas where blue boxes are functions that transform data and white boxes are datasets, linking each plot back to its sources. They use an Obsidian based skill that scans a repository, updates the canvas, and helps coordinate tasks with PyCharm and its coding agent to produce Jupyter or R notebooks. Over a conversation lasting about two or three days, the agent helped refine a linear model and discuss outputs using screenshots, while also supporting results by finding relevant papers."}