What problem does it solve? Roof measurement in a CRM often relies on unreliable global-mask calculations that silently produce wrong areas and can overwrite estimates. This Skill defines a perimeter-first implementation contract for the RoofTrace AI feature in PITCH CRM's Measure Lab, ensuring AI-assisted roof traces are calibrated, gated, reviewed, and saved as measurement drafts without ever auto-writing estimates. ## Core Features & Use Cases - Perimeter-First Tracing Workflow: Enforces a 7-stage pipeline (source, calibrate, outer perimeter, topology, pitch, review, approve) where the outer eave perimeter must be accepted before any facet, ridge, hip, or valley data is trusted. - Hard Quality Gates: Blocks approval when scale confidence is below 0.85, the perimeter self-intersects, facets fall outside the perimeter, or topology lines cross, and flags low-confidence penetrations or stale imagery. - Immutable Revisions & Measurement Drafts: Approval creates a versioned, immutable revision that writes only to measurement_drafts, never touching estimates, materials, or labor rows. - Use Case: An estimator uploads an aerial image for a job, calibrates scale from a known dimension, accepts the AI-proposed perimeter, edits facets, sets a 6/12 pitch, and approves a revision that produces exactly one measurement draft for later estimate building. ## Quick Start Load this Skill when implementing or modifying the RoofTrace AI tracing workbench, the roof_trace_sessions, roof_trace_revisions, roof_trace_jobs, or measurement_drafts tables, or any Measure Lab perimeter-first measurement logic.
