sheleg-design-skill 1.64.2 → 1.66.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +42 -0
- package/package.json +1 -1
- package/plugins/sheleg-design/.claude-plugin/plugin.json +1 -1
- package/plugins/sheleg-design/commands/sheleg-design.md +1 -1
- package/plugins/sheleg-design/skills/sheleg-design/COMPONENT_LAYER.md +14 -0
- package/plugins/sheleg-design/skills/sheleg-design/CREATIVE_DIRECTOR.md +6 -3
- package/plugins/sheleg-design/skills/sheleg-design/MOTION_DOCTRINE.md +5 -2
- package/plugins/sheleg-design/skills/sheleg-design/SKILL.md +8 -24
- package/plugins/sheleg-design/skills/sheleg-design/STYLE_PACK_INDEX.md +16 -0
- package/plugins/sheleg-design/skills/sheleg-design/VISUAL_REVIEW.md +28 -0
package/CHANGELOG.md
CHANGED
|
@@ -1,3 +1,45 @@
|
|
|
1
|
+
## 1.66.0 (2026-10-20): a short description, triggers in the body
|
|
2
|
+
|
|
3
|
+
- **The skill description is a router of 425 characters** (was 964). The
|
|
4
|
+
operator's Claude Code skill listing measured 29,995 of about 30,000
|
|
5
|
+
characters for 134 skills, and 81 skills showed no description; a plugin
|
|
6
|
+
skill cannot be hidden, so length is the only lever. One "Use when" sentence
|
|
7
|
+
that keeps every surface class and T1 carrier, five trigger phrases, a "Not
|
|
8
|
+
for" boundary naming `ux-flows` and `copywriting`, and the refusal phrase.
|
|
9
|
+
- **Every removed trigger phrase moved verbatim into a `## Triggers` section**
|
|
10
|
+
of the SKILL.md body, before *When to Use*, with the four bare nouns the hook
|
|
11
|
+
routes only behind an action verb.
|
|
12
|
+
- **The `/sheleg-design` command carries one sentence** (75 characters, was 213).
|
|
13
|
+
- **Two passages left the body so it stays inside the house auditor's 4,750-token
|
|
14
|
+
working limit** (4,747 before, 4,706 after, with the new section): *Who does the
|
|
15
|
+
work* moved to `COMPONENT_LAYER.md`, whose header already said it lived there,
|
|
16
|
+
and *What a comparison records* moved to `STYLE_PACK_INDEX.md` § Selection
|
|
17
|
+
protocol; `SKILL.md` links both. `fix-ds-01.02` reads the bundle. The row table
|
|
18
|
+
stays in `SKILL.md`, as `CALIBRATION.md` records.
|
|
19
|
+
- **`validate.py`:** the description is a single-quoted scalar so `Triggers:`
|
|
20
|
+
carries a colon rather than a spaced hyphen; the YAML plant now strips those
|
|
21
|
+
quotes. The Russian routing forms are checked in the `## Triggers` section,
|
|
22
|
+
not the description. A description over 500 characters or without "Not for"
|
|
23
|
+
fails, and a command description over 120 fails.
|
|
24
|
+
|
|
25
|
+
## 1.65.0 — a rendered surface is checked in WebKit (Safari) first
|
|
26
|
+
|
|
27
|
+
- **Engine order in the visual-review contract.** `VISUAL_REVIEW.md` gains
|
|
28
|
+
*Engine order — WebKit (Safari) first*: Safari through `safaridriver`,
|
|
29
|
+
Playwright's `webkit` (`--browser webkit`, `npx playwright install webkit`),
|
|
30
|
+
Safari Web Inspector, and `open -a Safari <url>` on macOS rather than the
|
|
31
|
+
system default browser. Chromium is a stated fallback for a task that needs
|
|
32
|
+
Chrome itself (a Chrome-only API, Lighthouse, a CDP trace, the chrome-devtools
|
|
33
|
+
MCP), on the Chrome Dev build, with a temporary profile and every started
|
|
34
|
+
process closed. A Chromium-only check is reported as exactly that.
|
|
35
|
+
- The capture record gains an `engine` field. Capture stays tool-agnostic:
|
|
36
|
+
any working browser capability still qualifies and Playwright is still not required.
|
|
37
|
+
- Motion review (`MOTION_DOCTRINE.md` §12) and the director's verify lane
|
|
38
|
+
(`CREATIVE_DIRECTOR.md`) put the WebKit route first and keep the Chrome
|
|
39
|
+
DevTools tooling as the stated Chromium fallback.
|
|
40
|
+
- New regression `test/audit_regressions/safari-first.py`, with mutations that
|
|
41
|
+
drop each point. Text integrity only; no claim that a model follows it.
|
|
42
|
+
|
|
1
43
|
## 1.64.2 — derived palette facts and explicit calm fallback
|
|
2
44
|
|
|
3
45
|
- Dark theme documentation now includes source-derived tables for almanac,
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "sheleg-design-skill",
|
|
3
|
-
"version": "1.
|
|
3
|
+
"version": "1.66.0",
|
|
4
4
|
"description": "Design taste as an installable agent skill. Cinematic scroll-driven landing pages built on one scroll clock and layered degrade-to-calm motion, a motion doctrine that decides whether to animate before it decides how, three calibration dials, and thirty-nine locked style packs with ready-made design tokens \u2014 instrument-console, editorial-luxury, workbench, briefing-room, atrium, orchard, outrank, babylove, patchbay, nameplate, field-notes, cyclorama, showroom, blueprint, prism, maquette, scoreboard, datasheet, manpage, pigeonhole, roster, ora, tenor, paperclip, ledger, router, daylight, notation, almanac, vitrine, proscenium, awning, bulletin, rimlight, onionskin, deskmate, test-drive, surveyor and chorus. Colour, slop and fork-reciprocity gates run as scripts, not opinions. Works with Cursor, Claude Code and any agent that reads a SKILL.md.",
|
|
5
5
|
"bin": {
|
|
6
6
|
"sheleg-design-skill": "bin/cli.js"
|
|
@@ -3,7 +3,7 @@
|
|
|
3
3
|
"name": "sheleg-design",
|
|
4
4
|
"displayName": "SHELEG Design",
|
|
5
5
|
"description": "SHELEG Design methodology: cinematic scroll-driven landing pages (single scroll clock, layered degrade-to-calm motion, WebGL particle formations), a motion doctrine that decides whether to animate before it decides how, and thirty-nine pluggable visual style packs, each with a ready-made token layer — instrument-console (near-black aerospace console; technical systems and infrastructure), editorial-luxury (cream and espresso; editorial, research, premium B2B), workbench (quiet light/dark UI; dashboards, admin, internal tools), briefing-room (dark 16:9 deck; investor and board presentations), atrium (cream daylight; consumer health and high-trust DTC), babylove (white with orange; SEO SaaS, long time-to-value, disconnected states), patchbay (black with mint-cyan; engines, buses, pipelines, OSS front doors), nameplate (cool slab with coral; press, certification and third-party trust), rimlight (white with blue light; studios, services and case studies), onionskin (white working sheet; developer and AI infrastructure), deskmate (warm beige with a dusk bleed; AI coworkers and chat-native agents), test-drive (warm paper with one coral; self-serve SaaS proven by the live product), surveyor (peach paper with teal and pink; visibility, monitoring and benchmark tools), chorus (warm paper under a crosshair grid, with a cut-corner bubble; AI-search visibility, brand monitoring and community marketing), outrank (white with violet; one brand across marketing and product), orchard (warm oat slabs; friendly biotech and wellness), field-notes (ruled green paper; auditable open-source developer tools), showroom (white gallery; product-led companies selling the app), blueprint (white technical stock; vector, storage and query infrastructure), prism (iridescent hard edge; command-first OSS infrastructure), maquette (near-black table; enterprise data infrastructure), cyclorama (looping pastel field; applied-AI consultancies), scoreboard (warm paper; growth, ads and accumulating metrics), datasheet (off-white specification; fraud, identity and device intelligence), manpage (cream manual; APIs, SDKs, CLIs and MCP servers), pigeonhole (filed white field; inbox, ticket and CRM triage), roster (faint square grid; platforms sold through who uses them), ora (warm coal; machine verdicts, audits and protocol traces), tenor (warm management paper; agent operations and autonomous back office), ledger (cream ruled console; analysts, BI and warehouse agents), paperclip (monochrome coal; orchestrators, schedulers and job runners), awning (white commerce forecourt; storefront, payroll and billing platforms), router (pale console with hairlines; dashboard, billing and inventory surfaces), daylight (bright portal with one shadow; client onboarding and service portals), notation (restrained technical hairlines; OSS, docs and developer workspaces), almanac (oatmeal editorial paper; manifestos and category-defining pages), vitrine (serif trust display; evaluated B2B, security and compliance), proscenium (white demo sequence; product tours and launch pages), bulletin (cheerful outlined bands; broad SMB and agency platforms). Ships the sheleg-design skill, the architecture reference, the motion doctrine, the Figma and Claude Design bridges, AI-surface patterns, the packs' token CSS, and the /sheleg-design command.",
|
|
6
|
-
"version": "1.
|
|
6
|
+
"version": "1.66.0",
|
|
7
7
|
"author": {
|
|
8
8
|
"name": "ssheleg",
|
|
9
9
|
"url": "https://x.com/sshlg93"
|
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
---
|
|
2
|
-
description: Apply
|
|
2
|
+
description: Apply SHELEG Design to the current landing, hero, motion or product UI task
|
|
3
3
|
argument-hint: "[what to build or upgrade]"
|
|
4
4
|
---
|
|
5
5
|
|
|
@@ -14,6 +14,7 @@ starter theme.
|
|
|
14
14
|
|
|
15
15
|
- [Why it composes rather than competes](#why-it-composes-rather-than-competes)
|
|
16
16
|
- [The two vocabularies, and the adapter contract](#the-two-vocabularies-and-the-adapter-contract)
|
|
17
|
+
- [Who does the work](#who-does-the-work)
|
|
17
18
|
|
|
18
19
|
## Why it composes rather than competes
|
|
19
20
|
|
|
@@ -41,3 +42,16 @@ component matrix (Card/Button/Dialog: computed sizes, padding, radius, shadow,
|
|
|
41
42
|
fonts, states) against the chosen direction**, not just the color variables, and
|
|
42
43
|
list the defaults you deliberately kept. A kit mounted without any of that
|
|
43
44
|
renders in its starter look and looks like nobody chose anything.
|
|
45
|
+
|
|
46
|
+
## Who does the work
|
|
47
|
+
|
|
48
|
+
Moved from `SKILL.md` on 2026-10-20, as this file's header already promised.
|
|
49
|
+
|
|
50
|
+
This section decides *whether* and *against which tokens*.
|
|
51
|
+
Where the `shadcn` skill is installed it does the adding, searching and
|
|
52
|
+
composing, and `migrate-radix-to-base` handles the Radix→Base move; both are
|
|
53
|
+
**optional neighbours** this skill declares in its `compatibility` line and does
|
|
54
|
+
not ship, and they trigger on their own words where present. Where they are
|
|
55
|
+
absent nothing is blocked: the kit's own CLI (`npx shadcn@latest add
|
|
56
|
+
<component>`) does the adding, and the token mapping above binds either way. Do
|
|
57
|
+
not restate their component docs here.
|
|
@@ -323,9 +323,12 @@ until the judge has been measured against a labelled set. The verdict goes into
|
|
|
323
323
|
the record's `ADA` field.
|
|
324
324
|
|
|
325
325
|
**Run them where the thing runs.** A screenshot in a browser at the target
|
|
326
|
-
viewport beats reading the diff, every time
|
|
327
|
-
|
|
328
|
-
|
|
326
|
+
viewport beats reading the diff, every time. WebKit first
|
|
327
|
+
([`VISUAL_REVIEW.md`](./VISUAL_REVIEW.md#engine-order)): Safari through
|
|
328
|
+
`safaridriver`, or Playwright's `webkit` under `webapp-testing`; the Chrome
|
|
329
|
+
DevTools tooling is the stated Chromium fallback when the task needs Chrome
|
|
330
|
+
itself. The `verify` lane in Act 2 casts them. A quality claim made from source
|
|
331
|
+
is a claim about source.
|
|
329
332
|
|
|
330
333
|
### Render critique — the soft half, made actionable
|
|
331
334
|
|
|
@@ -425,8 +425,11 @@ meant; the frames say what ships. Four steps, in this order, on the real build
|
|
|
425
425
|
1. **Capture a frame strip and a run video.** One frame per phase — start,
|
|
426
426
|
mid, settle; one per step of a stagger — not smooth playback, plus the video
|
|
427
427
|
of one full run.
|
|
428
|
-
- Web:
|
|
429
|
-
|
|
428
|
+
- Web: WebKit first ([`VISUAL_REVIEW.md`](./VISUAL_REVIEW.md#engine-order))
|
|
429
|
+
— a Playwright `webkit` video or a Safari Web Inspector timeline
|
|
430
|
+
recording; a Chromium DevTools performance trace with screenshots is the
|
|
431
|
+
stated fallback when the task needs the CDP trace itself. Frames are
|
|
432
|
+
pulled from it at the phase times.
|
|
430
433
|
- Native: the simulator's or device's screen recording
|
|
431
434
|
(`xcrun simctl io <device> recordVideo`, `adb shell screenrecord`).
|
|
432
435
|
- A rendered file ([`MOTION_PRODUCTION.md`](./MOTION_PRODUCTION.md)):
|
|
@@ -1,10 +1,10 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: sheleg-design
|
|
3
|
-
description: Use when deciding how something
|
|
3
|
+
description: 'Use when deciding how something looks or moves (landing, hero, particle/WebGL, scrubbed motion, drift, dashboard, admin and internal tool UI, mobile screen, web deck, tokens, light/dark theme). Triggers: "design a landing" / "дизайн лендинга", "palette" / "палитра", "redesign" / "редизайн", "мобильный экран", "figma variables". Not for structure (ux-flows) or copy (copywriting). Refusal phrase "no design" / «без дизайна».'
|
|
4
4
|
license: MIT
|
|
5
5
|
compatibility: Kit/lane commands need node>=16, npx, network. Optional siblings — dataviz, shadcn, migrate-radix-to-base. Each has an in-text fallback when absent or unreachable.
|
|
6
6
|
metadata:
|
|
7
|
-
version: 1.
|
|
7
|
+
version: 1.66.0
|
|
8
8
|
---
|
|
9
9
|
|
|
10
10
|
# SHELEG Design
|
|
@@ -56,6 +56,10 @@ unresolved observation stays unresolved rather than being renamed shipped. A gla
|
|
|
56
56
|
cannot produce a conformance row — a receipt where it did is refused, because
|
|
57
57
|
"looked accessible" is a screenshot claim wearing a criterion's name.
|
|
58
58
|
|
|
59
|
+
## Triggers
|
|
60
|
+
|
|
61
|
+
"build a landing page" / "сделай лендинг", "dashboard style" / "стиль дашборда, сделай дашборд, админка", "scroll animation" / "скролл-анимация, анимации", "design tokens, style pack" / "дизайн-токены", "light/dark theme" / "светлая/тёмная тема", "figma variables" / "переменные фигмы, фигма в код, макеты", "palette, colors" / "палитра, цвета", "typography, font" / "типографика, шрифт", "how it looks, make it prettier" / "выглядит, красиво, красивее", "visual reference" / "визуальные референсы", "investor deck as a web page" / "веб-презентация", "redesign" / "редизайн, свёрстай, вёрстка".
|
|
62
|
+
|
|
59
63
|
## When to Use
|
|
60
64
|
|
|
61
65
|
- Landing/marketing/hero pages where motion is a stated goal; particle or WebGL
|
|
@@ -269,19 +273,8 @@ property does not error, it silently falls back — and a comparison built on
|
|
|
269
273
|
silent fallbacks compares the fallbacks. No CSS-swap promise follows:
|
|
270
274
|
components keep consuming the pack's own tokens.
|
|
271
275
|
|
|
272
|
-
**What a comparison records
|
|
273
|
-
|
|
274
|
-
when everything else is pinned: the receipt per candidate names the pack, its
|
|
275
|
-
adapter status, the compared component's geometry (positions and sizes as
|
|
276
|
-
rendered) and its states (hover, focus, disabled). A content or viewport
|
|
277
|
-
mismatch between two records INVALIDATES the pair — it never counts as a pack
|
|
278
|
-
difference. A pack switch is *expected* to change rendered semantics — that is
|
|
279
|
-
what is being chosen — so identical geometry and states across a compared pair is a
|
|
280
|
-
signal the adapter never reached the component, not a finding of equivalence.
|
|
281
|
-
And a candidate without a complete adapter makes the whole comparison **not
|
|
282
|
-
ready**: the pack is named and either adapted or struck from the candidate
|
|
283
|
-
list explicitly — "compared what had adapters" silently is a comparison of a
|
|
284
|
-
different question.
|
|
276
|
+
**What a comparison records**, and when the whole comparison is **not ready**:
|
|
277
|
+
[`STYLE_PACK_INDEX.md`](./STYLE_PACK_INDEX.md#selection-protocol).
|
|
285
278
|
|
|
286
279
|
The comparison harness is scaffolding, not a deliverable: it comes out with the
|
|
287
280
|
same change that records the decision.
|
|
@@ -320,15 +313,6 @@ default. A scroll-driven landing page: **no** — there are no controls to reuse
|
|
|
320
313
|
there, the work is bespoke scroll and WebGL, and reaching for a component kit is
|
|
321
314
|
how a hero ends up looking like a settings screen.
|
|
322
315
|
|
|
323
|
-
**Who does the work.** This section decides *whether* and *against which tokens*.
|
|
324
|
-
Where the `shadcn` skill is installed it does the adding, searching and
|
|
325
|
-
composing, and `migrate-radix-to-base` handles the Radix→Base move; both are
|
|
326
|
-
**optional neighbours** this skill declares in its `compatibility` line and does
|
|
327
|
-
not ship, and they trigger on their own words where present. Where they are
|
|
328
|
-
absent nothing is blocked: the kit's own CLI (`npx shadcn@latest add
|
|
329
|
-
<component>`) does the adding, and the token mapping above binds either way. Do
|
|
330
|
-
not restate their component docs here.
|
|
331
|
-
|
|
332
316
|
## Optional — Figma (design ↔ code)
|
|
333
317
|
|
|
334
318
|
Touching a Figma file — publishing the pack as variables, or building from a
|
|
@@ -69,3 +69,19 @@ route cannot expose the difference.
|
|
|
69
69
|
|
|
70
70
|
The pack wins on values. The calibration dials in `SKILL.md` win on amount.
|
|
71
71
|
Neither may invent a token the chosen pack does not define.
|
|
72
|
+
|
|
73
|
+
Moved from `SKILL.md` on 2026-10-20 to keep its body inside the house budget.
|
|
74
|
+
|
|
75
|
+
**What a comparison records: one component, one content, one viewport at a
|
|
76
|
+
time, through the adapters.** A difference is attributable to the pack only
|
|
77
|
+
when everything else is pinned: the receipt per candidate names the pack, its
|
|
78
|
+
adapter status, the compared component's geometry (positions and sizes as
|
|
79
|
+
rendered) and its states (hover, focus, disabled). A content or viewport
|
|
80
|
+
mismatch between two records INVALIDATES the pair; it never counts as a pack
|
|
81
|
+
difference. A pack switch is *expected* to change rendered semantics (that is
|
|
82
|
+
what is being chosen), so identical geometry and states across a compared pair is a
|
|
83
|
+
signal the adapter never reached the component, not a finding of equivalence.
|
|
84
|
+
And a candidate without a complete adapter makes the whole comparison **not
|
|
85
|
+
ready**: the pack is named and either adapted or struck from the candidate
|
|
86
|
+
list explicitly, because "compared what had adapters" silently is a comparison of a
|
|
87
|
+
different question.
|
|
@@ -13,6 +13,7 @@ rather than dressed up as a reading.
|
|
|
13
13
|
- Fitness — what disqualifies a frame
|
|
14
14
|
- Worst-case data before the capture
|
|
15
15
|
- Capture is tool-agnostic
|
|
16
|
+
- Engine order — WebKit (Safari) first
|
|
16
17
|
- Visual evidence is its own class — do not launder others into it
|
|
17
18
|
- Rendered copy — the design/copywriting seam
|
|
18
19
|
|
|
@@ -33,6 +34,7 @@ screenshot with no record is an image, not evidence:
|
|
|
33
34
|
| `motion` | reduced or full — a still of an animated surface says which |
|
|
34
35
|
| `captured-at` | ISO-8601 UTC, so staleness is computable |
|
|
35
36
|
| `source` | the capability that took it (see below) — never assumed |
|
|
37
|
+
| `engine` | the rendering engine behind it — WebKit, Chromium or Gecko; a frame from one engine says nothing about another |
|
|
36
38
|
|
|
37
39
|
## Fitness — what disqualifies a frame
|
|
38
40
|
|
|
@@ -75,6 +77,32 @@ not fail because one specific tool is absent, and it does not install a tool to
|
|
|
75
77
|
manufacture a pass. If no capability is present, that is the `unverified` case
|
|
76
78
|
above, stated plainly.
|
|
77
79
|
|
|
80
|
+
<a id="engine-order"></a>
|
|
81
|
+
## Engine order — WebKit (Safari) first
|
|
82
|
+
|
|
83
|
+
Which tool takes the frame is open; which engine renders it first is not.
|
|
84
|
+
**Check a rendered surface in WebKit first.** Its support envelope is the
|
|
85
|
+
narrowest of the three big engines, so a pass there is the stricter claim.
|
|
86
|
+
|
|
87
|
+
| Need | WebKit route (first) |
|
|
88
|
+
|---|---|
|
|
89
|
+
| a real browser, driven | Safari through `safaridriver` (WebDriver); enable remote automation once with `sudo safaridriver --enable` |
|
|
90
|
+
| automation and captures | Playwright's `webkit` — the `webkit` project, `--browser webkit`, installed with `npx playwright install webkit` |
|
|
91
|
+
| debugging a render | Safari Web Inspector |
|
|
92
|
+
| opening a page by hand on macOS | `open -a Safari <url>` — not the system default browser, which may be another engine |
|
|
93
|
+
| an automation library (browser-use and the like) | its WebKit/Safari backend, where it has one |
|
|
94
|
+
|
|
95
|
+
**Chromium is a stated fallback, never the default.** Use it only when the task
|
|
96
|
+
needs Chrome itself — a Chrome-only API, Lighthouse, a CDP trace, the
|
|
97
|
+
chrome-devtools MCP — and say which and why in the record. Prefer the Chrome Dev
|
|
98
|
+
build for that run. When you launch your own Chrome or Chromium for a UI run,
|
|
99
|
+
give it a temporary profile (never the user's), record what you started, and
|
|
100
|
+
close every process you started before you finish.
|
|
101
|
+
|
|
102
|
+
A surface checked only in Chromium is reported as exactly that — `engine:
|
|
103
|
+
Chromium`, "checked only in Chromium" — never as "tested in a browser". Where
|
|
104
|
+
the claim is about cross-engine behaviour, both engines appear in the record.
|
|
105
|
+
|
|
78
106
|
## Visual evidence is its own class — do not launder others into it
|
|
79
107
|
|
|
80
108
|
Four kinds of evidence answer different questions, and one may not be relabelled
|