Solutions
AI workflows
Loop engineering for prompt optimizationA goal-driven generate–review–revise loop that turns style reverse-engineering and generalization testing into a reusable processBuilding a component libraryFrom shadcn source to Figma variables to Base UI + Storybook: semantic tokens on a headless foundation for a fully controlled component libraryDesign HarnessFrom a single design.md to a five-layer architecture of intent, source of truth, disclosure, constraints, and feedback, so agents ship on-brand UI reliablySkillManager: one source, many entriesKeep one canonical copy of each cross-agent skill in a central repo and syndicate it to every AI assistant via symlinks — edit once, apply everywhere
Skills
AGENTS-GENTreat AGENTS.md as scarce always-on context: every rule needs repository evidence, the right scope, and long-term valueSuperBrainA configurable, versionable coordination layer between a host agent and external CLIs, running two-round multi-seat deliberationgood-readme & GitworkLet AI wrap up after coding: keep the README in sync and commit only the changes from the task at handIMG to FigmaComplete process to turn UI images into editable design drafts