What problem does it solve? Turning discovery-layer judgment (intents, insights, journeys, blueprints) into a buildable specification is error-prone: requirements get invented, ambiguities get silently resolved, and traceability is lost. This Skill derives a complete, traceable requirement set in one pass so a downstream executor can build without guessing. ## Core Features & Use Cases - Traceable requirement derivation: Every SHALL cites its earning source via a traces_to grammar (settled proposals, supported insights, moments of truth, boundary rules), and unproven ground routes to RESEARCH documents instead of becoming requirements. - Two input modes: Intent-only mode for system work, and service-design mode that additionally consumes ecosystem, persona, journey, blueprint, and Discovery documents. - Escalation over invention: Contradictions and gaps are escalated to the caller rather than resolved in-body, and missing or schema-invalid inputs park the run. - Instrumentation: Emits uncertainty predictions for set coverage and each RESEARCH hypothesis, witnessed later by the crit step. - Use Case: Given a settled intent document for a CLI surface, generate the full FR/NFR set with typed ids (FR-CMD-001, NFR-CMD-001), validate samples against the store schemas, batch-create the documents, and report escalations. ## Quick Start Ask the agent to write the spec for a scope by naming its intent document and surface, for example: write the requirement set for intent INT-AUTH on the auth surface.