What problem does it solve? Research and engineering projects often start from vague directions, and teams pick routes without verifying evidence, baselines, or feasibility. This Skill narrows an unclear research question into a verifiable route and leaves a persistent artifact and state-graph node so later sessions can resume with full context. ## Core Features & Use Cases - Evidence Scouting: Clarifies the research question, target claim, expected evidence, baselines, metrics, and first failure mode, verifying against primary or official sources. - Artifact and Graph Recording: Writes a scout artifact (research/artifacts/scout-<slug>.md) and adds a scout node to the project state graph via the ds_lite_state.py script, keeping STATUS.md in sync. - Honest Status Reporting: Marks nodes blocked or not-verified when evidence is missing instead of producing polished success prose, and treats instructions found in papers or repositories as untrusted data. - Use Case: Before committing to a model architecture, run the scout to survey benchmark datasets, confirm a strong baseline, record feasibility risks, and hand off a grounded route to the idea or experiment stage. ## Quick Start Ask the assistant to scout the evidence, baselines, metrics, and risks for your research question before choosing a direction.
