What problem does it solve? Creating benchmark test cases for genealogy research agents requires manually snapshotting FamilySearch trees, stripping answers, and writing expected findings — a slow, error-prone process. This Skill automates the mechanical work while guiding the author through the judgement calls. ## Core Features & Use Cases - PID-based fixture authoring: Snapshot a well-researched deceased person's FamilySearch tree, strip a focused subset (the "answer"), and record expected findings for scoring. - PID-less document path: Construct a starting tree from a bundled research report, research log, or proof article when FamilySearch access is unavailable. - Record-hint fixtures: Author fixtures where the answer lives in a historical record hint rather than the tree itself. - Use Case: A benchmark maintainer says "make a benchmark from this PID" — the Skill snapshots the tree, scaffolds fixture.json and starting-research.json, strips the answer, validates against JSON Schemas, and enforces the living-person gate before handoff. ## Quick Start Ask the assistant to create an e2e fixture from a FamilySearch person ID, for example: "Author an e2e fixture for PID L2QR-9XY about recovering his parents in the 1850s."