What problem does it solve? Research tasks can pass software tests while still being scientifically wrong — incorrect formulas, null models, topology semantics, estimands, provenance, or paper claims. This Skill gives managers a triage procedure to classify tasks by assurance lane and attach machine-checkable or human-review requirements before dispatch. ## Core Features & Use Cases - Assurance Lane Classification: Sorts tasks into six lanes — topology, stochastic/null model, statistical/panel, representation, output/provenance, and paper claim. - Dispatch and Review Checklists: Provides a Task Prompt block, worker evidence expectations, and a manager review checklist covering seeds, parameters, schemas, and caches. - Lane Routing Table: Maps each lane to judgment skills and deterministic enforcement contracts for follow-up validation. - Use Case: Before dispatching a task that recomputes permutation p-values, a manager uses this Skill to identify the stochastic-null lane, require a Monte Carlo formula contract, and demand provenance fields in the output JSON. ## Quick Start Ask the assistant to triage a planned research task using the research-assurance-triage skill and produce the Research Assurance Requirements block for the task prompt.
