Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add pehcastro/harness --skill nkz-tastegit clone --depth 1 https://github.com/pehcastro/harnessWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/pehcastro/harness/nkz-taste)<a href="https://agentmods.dev/skills/pehcastro/harness/nkz-taste"><img src="https://agentmods.dev/badge/skills/pehcastro/harness/nkz-taste/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/pehcastro/harness/nkz-taste"><img src="https://agentmods.dev/badge/skills/pehcastro/harness/nkz-taste.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00147 | $0.03435 |
| Opus 5.5 | $0.00059 | $0.01374 |
| Sonnet 5.5 | $0.00029 | $0.00687 |
| Haiku 4.5 | $0.00015 | $0.00344 |
Grade A, and why
nkz-taste scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 10d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 180 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Taste
Taste is the art of the invisible. You look at it and you know it is good, and you cannot point at the one thing that made it so.
Four layers, in this order. Each one is a different question, and a good answer to a later question cannot rescue a bad answer to an earlier one.
| Layer | Question | Where |
|---|---|---|
| Product | Should this exist, and what is its one job? | product.md |
| Craft | Does it work for everyone, in every state? | craft.md |
| Direction | What should it look like, and why this and not the default? | direction.md |
| Feel | Does it look, behave, and move well? | design / interaction / animation files |
Most UI work fails at Product and gets polished at Feel. A beautiful screen that does the wrong job is a more expensive mistake than an ugly screen that does the right one. Do not start at Feel.
Within Feel:
Design → how it looks
Interaction → how it behaves
Animation → how it feels
Two rules hold across all of it:
- Nobody should notice an animation. If a user notices motion, it was too much. Motion is for continuity and feedback, never for attention.
- Interaction without feedback is not interaction. Every action confirms itself inside 100ms, before the work is done.
The loop
Run this in order. Do not skip step 1 because the request sounds small: a request for one button still sits inside a screen that has a job.
- Frame the product. Name the surface, the user's job, the one primary action, and what happens before and after. Identify unnecessary additions, if any. Read product.md. Ten minutes here changes what you build; ten minutes at step 6 changes a shadow.
- Survey what is in place. Find the token source, the component library, the animation library, the existing conventions. See In place vs out of place below.
- Resolve the style. If a style was selected, load its contract and recipes using the router below; those decisions are already made. Otherwise set the direction, when the work is new or brand-facing. State the read, set the intensity dials, write the four-part plan, then critique it against the answer you would give any similar brief. Read direction.md. Skip this only when you are working inside an established system that already made these decisions.
- Pick the platform and load the files. Native mobile and web share philosophy and share almost no mechanics. See the router below.
- Propose before you build, when the change is more than a component. State the frame, the direction, the layout, and the trade-off you are making. One paragraph, not a document.
- Build, installing the selected style centrally when applicable, then taking every visual value from the project's token source. Never invent a hex, a font, or an arbitrary px inline. See Tokens below.
- Pass the pre-flight. Blocking. See below.
- Report the trade-offs. Name what you knowingly skipped and why, and name every new token the project needs.
What ships with it
12 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- README.md 1.5 KB
- references/ai-tells.md 15 KB
- references/animations.md 21 KB
- references/craft.md 14 KB
- references/direction.md 14 KB
- references/interactions.md 8.0 KB
- references/laws.md 7.1 KB
- references/mobile-design.md 14 KB
- references/product.md 13 KB
- references/web-design.md 8.4 KB
- styles/recess/recipes.md 12 KB
- styles/recess/style.md 16 KB
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 10d ago First seen · 180 lines · 147 tokens per session scan A e97086b2f079
nkz-taste is a skill published in the GitHub repository pehcastro/harness (1 stars, last pushed 11d ago), licensed MIT. It adds 147 tokens to every session and 3,435 once invoked, about $0.0006 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-28.
Other skills, from other repositories
ios-visual-review
Use when the user invokes /ios-visual-review to audit an iOS app's visuals on a real device. Not for remote, credential, publish, deploy, or irreversible changes.
design-android
Use when designing an Android screen or flow — deliverable is a device-framed HTML mockup and developer handoff spec, never Compose code.
design-ios
Use when designing an iOS or macOS screen or flow — deliverable is a device/window-framed HTML mockup and handoff spec, never SwiftUI code.
icon-composer-app-icon-workflow
Design, preview, validate, and integrate Apple app icons with Icon Composer. Use when a task mentions Icon Composer, .icon documents, Liquid Glass app icons, ictool preview exports, app icon design for iOS, iPadOS, macOS, or watchOS, or guided Icon Composer GUI work through Computer Use.
apple-typography-workflow
Guide Apple platform typography decisions and implementation across SwiftUI Font, UIKit UIFont, AppKit NSFont, San Francisco system fonts, SF Pro, SF Compact, SF Mono, SF Pro Rounded, New York serif design, Dynamic Type, text styles, fontDesign, fontWeight, fontWidth, monospaced digits, custom font bundle integration…
apple-ui-accessibility-workflow
Guide Apple UI accessibility implementation and review for SwiftUI-first code, with UIKit and AppKit bridge guidance plus explicit verification expectations and testing handoffs. Use when the user wants help making Apple UI accessible, reviewing accessibility semantics, shaping the accessibility tree, or planning…