codex-lane-worker
Bounded Codex Lane writer protocol for one assigned task. Use only when the lane runtime assigns this worker.
Context-first anti-slop visual review for landing pages, portfolios, and existing web UI. Audits evidence, hierarchy, rhythm, assets, states, and purposeful motion. Adapted from Leonxlnx/taste-skill. Use when: UI слоп, нейрослоп дизайн, taste, иерархия, воздух, Inter, фиолетовый
$ npx -y skills add VKirill/claude-lane-stack --skill design-taste --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/design-tasteContext preview
The summary Claude sees to decide when to auto-load this skill.
Context-first anti-slop visual review for landing pages, portfolios, and existing web UI. Audits evidence, hierarchy, rhythm, assets, states, and purposeful motion. Adapted from Leonxlnx/taste-skill. Use when: UI слоп, нейрослоп дизайн, taste, иерархия, воздух, Inter, фиолетовый
name: design-taste description: "Context-first anti-slop visual review for landing pages, portfolios, and existing web UI. Audits evidence, hierarchy, rhythm, assets, states, and purposeful motion. Adapted from Leonxlnx/taste-skill. Use when: UI слоп, нейрослоп дизайн, taste, иерархия, воздух, Inter, фиолетовый градиент, три карточки. SKIP: DESIGN.md extraction (project-design), gray prototypes (page-prototype), product code (writer via web-design), Russian copy (ru-check)." argument-hint: "[info]" license: MIT
Lane adapter of [taste-skill](https://github.com/Leonxlnx/taste-skill). It reviews landing pages, portfolios, and redesigns, and guides isolated branded mockups. It does not write product UI, and it does not turn cabinets, admin screens, or data tables into landing pages.
If `$ARGUMENTS` is `info` / `справка`, print this block verbatim, then stop:
design-taste — визуальный аудит и направление для изолированного макета. 1) Design read: kind / audience / job / vibe / evidence from the brief, screen and DESIGN.md. 2) Three dials: DESIGN_VARIANCE / MOTION_INTENSITY / VISUAL_DENSITY. Existing tokens win. 3) Evidence audit: type, color, layout/rhythm, assets, states, motion, a11y. 4) Preflight: every finding has evidence, impact and an owner. Агент: design-lead MODE=audit или MODE=mockup Чинить: UI-ран после «делай» Роутер: /lane-stack:web-design info
Read the brief, requested route or screenshot, relevant `DESIGN.md`, current components, and the project's stack before judging taste. Identify:
Write one line: `Reading this as: <kind> for <audience>, serving <job>, with a <vibe> language, using evidence from <paths/screenshots>.`
For an existing UI, describe what is already true before proposing change. Preserve information architecture, working accessibility patterns, and the project's brand unless the brief asks for a change. If evidence is missing, write `DEGRADED: <missing evidence>` and make the smallest safe recommendation. Ask one question only when two plausible design reads would lead to different work.
Use the existing `DESIGN.md` reading when it exists. Otherwise choose a starting range from the brief; these are review dials, not quotas:
| Surface | VARIANCE | MOTION | DENSITY | |---|---:|---:|---:| | Marketing / landing | 7–8 | 5–6 | 3–4 | | Portfolio / editorial | 6–8 | 4–7 | 2–4 | | Cabinet / admin / operate | 3–4 | 2–3 | 5–7 | | Trust / a11y-first | 3–4 | 2–3 | 4–5 |
Low variance favors a stable grid; high variance can justify asymmetry only when it helps the page job. Motion and density must follow the same reading.
Report findings as `finding → evidence (path, line, screenshot, or observed state) → impact → recommendation → owner`. Classify by consequence: blocker (primary task cannot be completed), major (material usability/accessibility barrier), or minor (polish). Do not call a pattern slop merely because it is common; explain why it fails this page's job.
Review these headings:
number alignment, wrapping, and whether the existing typeface fits the brand.
border/shadow logic, and consistent lighting. Inter is valid when the brand or product already uses it; do not swap fonts for novelty.
pacing, baseline alignment, and a spacing scale that comes from the project tokens. Use cards only when elevation communicates hierarchy.
crops, meaningful alt text, and no fake screenshots or decorative assets that compete with the job. Record an asset gap instead of inventing a replacement.
current navigation state where the surface needs them. Check keyboard behavior.
or a state change. Match the motion dial, use the project's existing tools, and provide a `prefers-reduced-motion` path. If the reason cannot be stated in one sentence, remove the animation.
placeholder copy and imagery presented as finished content.
Do not copy arbitrary font bans, palette hex lists, word or cell-count quotas, GSAP skeletons, or upstream framework defaults. Follow this project's stack, tokens, accessibility baseline, and existing icon family.
Before handing off an audit, confirm:
Claude Lane Stack is a self-hosted toolkit for coordinating coding agents through CLI processes, task contracts, Git workspaces, and verification receipts (bin/run-controller:117-159, schemas/task-v2.schema.json:1-60).
Bounded Codex Lane writer protocol for one assigned task. Use only when the lane runtime assigns this worker.
Operate Claude Lane Stack schema-v2 runs from Codex. Use for lane status, start, resume, review, verification, evidence, or…
Детекция и коррекция AI-сгенерированного текста по методике DrMax LinguaForensic v3.9.4 (284 лингвистических признака + 16…
Owner-facing architect for a new app or service. Plain-language chat; living plan artifacts on disk. Use when user says…
Live browser QA + Playwright-compatible replay. Disk .agents/qa/. Use when (terminal Lane Stack): проверь в браузере, кликни…
Delegate fat file reads to the project's pm_read worker. Use when a Read/cat/head/tail/sed of a file over pm_read.min_lines was…