AI-Things/

Built for agents

Skills and prompt templates so LLMs ship Malevich-correct interfaces on the first try — all grounded in the v0.2.0 architecture (five component layers, three token tiers, seven modifier categories).

Skills 7 procedural skills — component usage, audit, refactor, authoring, foundations, modifiers, canvas import. Load one before a Malevich task.

Prompts 6 vetted prompt templates — generate a component or screen from a brief, refactor, audit, extend with modifiers, document a component.

MCP server · soon @malevich/mcp — tokens, component docs, and the modifier applicability matrix over MCP. Agents query "how to build a login form" and get the right components.