Muse Skill — Design skill for Claude Code
A Claude Code skill for design direction that gets past the obvious idea and stays inside your design system.
How to install Muse Skill
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open ikstro/muse-skill and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Muse Skill does
A Claude Code skill for design direction that gets past the obvious idea and stays inside your design system.
Alternatives in Design
- Brand Guidelines — Apply Anthropic's official brand colors and typography to artifacts 94.1k ★
- Od Contribute — Open a first-contribution PR (or bug issue) on nexu-io/open-design — works for non-coders too 91.8k ★
- Export Diagram — Export a diagram-design HTML file to .svg and .png next to the source 27.2k ★
README
Muse
**A Claude Code skill for design direction that gets past the obvious idea and stays inside your design system.**
Ask a model for "something more original" and it reaches for the cheapest levers: it changes the colours and the typeface, or it dresses the page up as a ticket, a map or a dashboard. Muse closes those levers. Your brand's materials stay locked, the category's clichés are crossed out before the first idea, and the only thing left to move is the idea itself: what the page shows, what happens in the first three seconds, and why the visitor wants the product afterwards.

*One of four motion comps Muse produced for Lanework, a fictional task tracker. The design system is the product's own; the idea is new.*
Why Muse gets past the obvious
Muse runs six mechanisms that keep the work from sliding back to the usual answer.
**The usual is named and excluded before the first idea.** Muse writes down what the category always ships and its predictable opposite, and crosses both out. They are not bad; they are taken. *In the example:* a screenshot in a browser frame over a gradient blob, "All your work in one place", three feature cards; and the opposite, a black terminal page. The category default scored 5 out of 25.
**Width, not variations of one thought.** The list is long and mixed: 20–25 directions across ten idea types (product as object, transformation, demonstration, moment of use, tension, a line of copy, a scale fact, a social moment, a reveal, and a metaphor last), at most four per type. *In the example:* 21 directions across all ten types.
**Randomness from outside.** A model is predictable: given the same task it keeps shipping the same top concept. So about a quarter of the list is built from an external random string that picks an idea type, a moment in the buyer's life, a product fact and one constraint (for instance "the hea
Related Skills
Managoat Runtimes
How a coding agent CLI (claude, codex, gemini, opencode) gets into a sandbox and comes up speaking ACP: the ad
Shalva Inbox Agent
Shalva – a personal inbox assistant that learns how you write and file email, then gets your inbox down to ope
Icm Architect
Claude skill: design any process, idea, or problem into an ICM workspace (folder structure as agent architectu
Hef.Brainstorm
Socratic design exploration before specification — refine the idea before committing to requirements
Venture Assessor Skill
An Anthropic Claude Skill for evidence-first startup idea evaluation, market research, competitor mapping, bea
Plan Stages
Bootstrap a staged-rollout .plan/ from a project idea — design, decompose, scaffold, and commit. Executes no s
Related Agents
Engine Design Review Architecture
Before a change you've asked for gets built, checks whether the plan is soundly designed — clean boundaries, a
Tier3 Complex
Tier 3, multi-step reasoning inside a fixed direction. Use when the approach or the question and the acceptanc
Eureka
Use this agent when Claude is playing it safe, stuck looping on wrong hypotheses, or when standard fixes aren'