What problem does it solve? One-off R Markdown analyses at FGCZ produce outputs that cannot be chained as parent datasets by downstream SUSHI apps and leave no trace in the B-Fabric audit trail. This Skill promotes a custom analysis into the exact on-disk folder contract every SUSHI app honours, making it visible in the SUSHI lineage tree and UI. ## Core Features & Use Cases - SUSHI-shaped scaffolding: Generates the timestamped {order_id}_{analysis_name}_{timestamp} folder with dataset.tsv, parameters.tsv, and input_dataset.tsv using the correct SUSHI column-type tags like [File,Link]. - Reproducible launcher: Emits a vanilla bash script that renders the Rmd via R --vanilla --slave, captures stdout/stderr logs, and copies the result to gstore with g-req. - Dual registration: Registers a B-Fabric workunit and dataset via register_custom_analysis.py, then inserts the production SUSHI data_sets row directly via MySQL with correct parent_id lineage and Ruby hash-rocket samples.key_value syntax. - Use Case: You rendered a QC Rmd consuming a FeatureCounts output for project p40992 and the user wants it delivered to gstore, chainable by ScSeurat, and visible at fgcz-sushi.uzh.ch/data_set/.... ## Quick Start Ask the agent to register your rendered R Markdown analysis as a SUSHI dataset for a given project and order ID, providing the upstream SUSHI dataset path and ID.
