committed-aesthetic
Skill112 starsChanged 2 months agoMITView on GitHub
How to write — and how to use — a skill that IS one aesthetic rather than a catalogue of them. A catalogue lets an agent pick, and it picks the modal option; a committed aesthetic makes it execute one thing precisely, against rules you can check. Use when a design keeps coming out competent and forgettable, when starting a new brand surface with no reference, or when authoring a house style you want executed the same way twice.
In this repository35 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- aesthetic-instrumentSkill
- anti-patternsSkill
- anydesignSkill
- archetype-review-baseSkill
- brainstormingSkill
- build-speedSkill
- codex-hostSkill
- cost-modelSkill
- crystallizeSkill
- decision-evalSkill
- deploy-landedSkill
- discoverySkill
- done-blockedSkill
- great_ctoSkill
- lifecycle-messagingSkill
- migration-ready-schemaSkill
- observability-baselineSkill
- opportunity-solution-treeSkill
- outcome-roadmapSkill
- pm-planningSkill
- pre-mortemSkill
- product-economicsSkill
- prose-styleSkill
- quant-validationSkill
- secrets-rotationSkill
- signing-preflightSkill
- skeptical-triageSkill
- stack-baselineSkill
- test-strategySkill
- ui-ux-pro-maxSkill
- vertical-onboardingSkill
- verticalsSkill
- well-architectedSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 111
If a system already governs, this skill does not apply — match what exists.
- Distinctive: line 112
If nothing governs and you can ask: offer 2–4 genuinely different directions, each an axis you can name, before building.
- Distinctive: line 115
If nothing governs and you cannot ask: commit to ONE, state the assumption in a line at handover, and put 1–2 low-fi alternates beside the …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
A committed aesthetic beats a catalogue
The problem this exists for
ui-ux-pro-max ships 84 styles, 161 palettes and 57 font pairings, and
design-advisor is wired to all of it. The catalogue is not the bottleneck.
The bottleneck is that a catalogue lets a model pick, and an unconstrained pick is the modal one. Asked for a landing page with no brand, any model reaches for the same safe centre: a soft gradient, a large sans heading, three rounded cards with a left accent border, generous neutral grey. Every input in the training distribution votes for it. Nothing in a catalogue entry saying "Swiss minimalism — clean, grid-based, generous whitespace" is specific enough to override that, because that sentence describes the safe centre too.
Compare a rule from a committed aesthetic:
every dimension is
calc(fraction * 100vw)against a baseline width, so one page at 1024px reads as a perfect scaled-down of the same page at 1920px
You can check that. A design either obeys it or does not. That is the whole difference: a catalogue entry is a label, a committed aesthetic is a constraint.
The form
A committed-aesthetic skill has five parts. The first is the one that does the work.
1. Transferable DNA — numbered, specific, falsifiable
Between 8 and 14 rules. Each must be checkable by someone who was not in the room. Test every rule against this: could two designers disagree about whether a page obeys it? If yes, it is a vibe, rewrite it.
| Vibe (useless) | Constraint (usable) |
| "thin, elegant type" | "weights 100 or 300 only, never 400+" |
| "lots of whitespace" | "~60% of any above-the-fold area is empty" |
| "restrained motion" | "no shadows, no gradients, no bounce; one shadow exists in the system, on toasts" |
| "technical feel" | "product codes use en-dash U+2013, never hyphen" |
| "dark and dense" | "five surface steps, each ~5% apart; depth is a surface step, never a shadow" |
2. What this does NOT do — the IP boundary, as a table
An aesthetic is inspired-by, not cloned. Name the specific assets that belong to whoever you are borrowing from — wordmark, icon set, information architecture, product names, copy — and for each give the reason and what to do instead. This is not a legal disclaimer at the bottom; it is a working table an agent reads while building, and it is the difference between an homage and a trademark problem. Where the source is a real company, default to inspired-but-distinct and ask before taking the clone path.
3. A component library that runs
One HTML file the agent can open, with every component rendered and its own CSS inline, copy-pasteable. A token table in prose is read and paraphrased; a rendered gallery is copied exactly. This is why the form works at all.
4. Templates for the signature patterns
The two or three moves nobody would reconstruct from the DNA list alone — the
navigation shape, the section header, the hero. With swap these for your own
brand marked in comments.
5. Reference captures, marked "study, do not pixel-copy"
Screenshots are for understanding the patterns. Building FROM a screenshot rather than from the rules produces a bad tracing of one page instead of a design in the family.
The banned list, and why we needed one
ui-ux-pro-max's anti-pattern tables are about accessibility, touch targets and
performance. They are good and they are not about taste. Nothing in this repository
told an agent what cliché looks like, so the only defence against slop was one
sentence in design-advisor: "no generic clean modern minimal filler".
Banned by default on any surface without an existing system that calls for them:
- A gradient as the page ground. Especially purple-to-blue, especially at 135°.
- The card triplet: three equal cards, rounded corners, a coloured left border, an icon in a tinted circle.
- Inter, Roboto, Arial, Open Sans, Fraunces, Poppins as the display face. Not because they are bad — because they are the default vote.
- Emoji as iconography. Draw the icon. (Emoji are fine if the brand uses them.)
- The hero that describes a category rather than taking a position. "The modern platform for X" is the sentence a model writes when it has not been told what to believe.
- Glassmorphism, neumorphism, and a mesh-gradient blob unless the brief names one of them.
- Symmetric three-column feature grids as the default second section.
A banned list is not a style. It removes the centre so a choice has to be made.
Using one
- If a system already governs, this skill does not apply — match what exists.
- If nothing governs and you can ask: offer 2–4 genuinely different directions, each an axis you can name, before building. Not five shades of one aesthetic — that is not a choice.
- If nothing governs and you cannot ask: commit to ONE, state the assumption in a line at handover, and put 1–2 low-fi alternates beside the deliverable, never instead of it.
- Once a direction is chosen, load its committed-aesthetic skill and execute the
DNA rules literally. Cite the rule number for each choice, the way
design-advisoralready cites a ui-ux-pro-max rule.
Instances in this repository
skills/aesthetic-instrument/SKILL.md— great_cto's own, measured from the board and greatcto.systems rather than designed for this document.
Related: skills/ui-ux-pro-max/SKILL.md is the
catalogue this stands beside, not a replacement for it — the catalogue is where
you look when you need to KNOW something; this is what you load once you have
DECIDED something.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

