Generate visuals across 11 channels including raster, SVG, web, interactive, video, terminal, diagrams, and infographics while enforcing brand tokens, 8:1 contrast minimum, custom dimensions, and style guides. Critique produced PNG/JPG/SVG/PDF artifacts with a vision-model agent that evaluates design principles (Tufte/Bertin/Gestalt), channel anti-patterns, and brand compliance, providing evidence-based PASS/NEEDS REVISION/FAIL verdicts.
Use this agent when creating visual narratives, designing infographics, building presentations, or communicating complex ideas through imagery. This agent specializes in transforming data and concepts into compelling visual stories that engage users and stakeholders. Examples:\n\n<example>\nContext: Creating app onboarding illustrations
A virtual design team for Claude Code, Cursor, Windsurf, Gemini CLI, and Copilot — powered by Naksha. Assembles specialist roles — UI designer, UX researcher, content designer, Figma expert, data viz, email, social, motion, presentation, brand strategy, illustration, video, conversational, spatial, compliance, and more — for any design task. 26 roles, 60 commands, 13,800+ lines of expert design knowledge. Your agency's design brain, inside your terminal.
Research-backed visual design principles for websites, presentations, documents, and any visual medium — grounded in VisAWI, Gestalt psychology, and empirical aesthetics research
Create beautiful, self-contained HTML visualizations from any content or idea. Use for slide decks, dashboards, infographics, flowcharts, timelines, cheat sheets, posters, and more.
muriel is a multi-constraint solver for visual production — a dozen channels of tool-use recipes for LLM agents, grounded in the design-history lineage they inherit. Ten output channels (raster, SVG, web, interactive, video, terminal, density viz, gaze, science, infographics) plus two cross-channel references (dimensions, style-guides), a two-tier brand-token schema with motion, anti-patterns per channel, live 8:1 contrast and OLED-palette enforcement at render time, and a vision-model critique agent grounded in Tufte / Bertin / Gestalt / Reichle / scanpath research.
The tooling is LLM-native: skill format, vision-model critique, brand tokens alive at render time, motion as a first-class schema field, engine adapters for Pillow / Flux / pretext / ffmpeg / Playwright. The principles are older: Cooper's Visible Language Workshop (tribute below), Tufte's data-ink discipline, Bertin's retinal variable ranking, Gestalt grouping, CRAP layout, Reichle's E-Z Reader, scanpath patterns from vision science. The tools serve the principles.
A single skill file (SKILL.md) teaches a Claude Code agent to generate every visual artifact a researcher-designer-engineer ships — from text source files that diff in git and regenerate from data. The constraint discipline (8:1 contrast, OLED palette, one font treatment, generated > drawn, reproducible > one-off) stays live at render time: brand tokens are parsed, contrast is audited, dimensions are enforced — not as lint after the fact, but as part of the act of making.
Heir projects — swap in your favorites
muriel is the grandmother to marginalia (editorial callouts and magazine layouts, cited throughout channels/web.md) and iblipper (kinetic typography and emotional-vocabulary animation, cited in vocabularies/kinetic-typography.md). Both grew from the same constraint discipline and ship as the defaults here because they're tuned to pass muriel's rules out of the box.
They're defaults, not requirements. The constraint discipline — 8:1 contrast, OLED palette, one font treatment, brand tokens live at render time — is the backbone. The specific libraries are preferences. Swap in your favorite editorial library, kinetic-typography engine, chart renderer, style-guide loader, imagegen provider, or rasterizer; muriel's opinions are about what constraints hold, not which library enforces them. Every channel doc names which library it assumes, and none of those assumptions are load-bearing against a sensible substitute.
Built on / integrates with
Python channels
Editorial
Interactive / graphics
Diagrams / video
Channels
Eleven output channels, each with its own subfile under channels/:
