vision-smith

name	vision-smith
description	Builds the smallest grounded visual unit for image/UI work: element-level screenshot descriptions, token-locked icons, or single-asset contrast fixes. Activates on /vision-smith or briefs that need one correct visual deliverable with observation separated from inference. Differentiator: token-locked build units with explicit seen-vs-inferred grounding.
argument-hint	<brief | path | artifact>
user-invocable	true
metadata	{"short-description":"Multimodal/vision · build unit","category":"vision","tier":"core","sg_id":"sg-0139","binary_id":"opgrok.sg.vision-smith","version":"3.1.0","leslie_gate":"v1","ai_polished":true,"production":true,"enriched":true,"intent":"vision/smith (build unit): Describe a UI screenshot with element-level grounding; Generate an icon set matching brand tokens; Edit an asset to fix contrast while keeping identity.","purpose":"Interpret or produce visual artifacts with grounding. Method (build unit): build the smallest correct unit that meets the brief. Domain: image/UI understanding and visual artifacts.","intent_tags":["vision","smith","core","build-unit"],"path":"core/skills/vision/smith/SKILL.md","call":"/vision-smith"}

Multimodal/vision Builder (/vision-smith)

Agent Identity: Hattie-a9b2427303ebd6cd4fd7c89dea3fcad779326e60f5f0e75e2c6edbd537858a84