What problem does it solve? Turning a set of visual references into a repeatable image style is error-prone: references conflict, approvals get inferred from silence, and style packages drift without evidence. This Skill provides a controlled calibration workflow that converts references, comparison specimens, and explicit decisions into a validated, portable visual-design-language (VDL) skill package. ## Core Features & Use Cases - Controlled calibration rounds: Hold fixture content, hierarchy, and geometry fixed while varying one style variable family at a time, with explicit approval recorded per specimen. - Reference role model and decision ledger: Classify every image as style-source, structure-source, anti-reference, or approved lock, and track decisions, conflicts, and uncertainties with evidence IDs. - Package generation and validation: Scaffold a complete VDL skill with init_vdl_skill.py, then enforce structure, identity, checksums, fixtures, and approval metadata with two validators before proposing release. - Use Case: A designer wants a reusable marker-wireframe style. They supply reference photos, run controlled specimen rounds to lock palette and mark-making, generate the package, and validate it with forward tests before release. ## Quick Start Use the calibrate-visual-language skill to turn my reference images into a validated reusable visual style package.
