CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

claude-design

Make and revise designs in the Design (canvas) artifact format (project/canvas.json plus one .dc.html Design Component file per artboard) as local files, then open them in a local canvas editor where the user can view the artboards side by side, retype copy and adjust tweaks. Use when the user asks for an editable canvas of artboards in this format, names claude-design or the Design (canvas) type, or wants to revise a folder that holds project/canvas.json. Not for a one-off HTML prototype or a design exploration outside this format.

Quality

91%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

The canonical home for this skill is claude-design in JimLiu/baoyu-design

SKILL.md
Quality
Evals
Security

Quality

Content

85%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body is exceptionally dense and actionable, with exact commands, file shapes, a complete example, and validated multi-step workflows. Its main weaknesses are missing referenced reference files (progressive disclosure is a promise the bundle doesn't keep) and over-detailed editor-behavior narration in Sharing/History.

Suggestions

Ship the ten referenced files (format.md, craft.md, diagrams.md, print.md, questions.md, view-state.md, design-system-components.md, brand-colors.md, brand-typography.md, make-from-this.md) under artifact-type/reference/, or inline their essential rules — as delivered, the body delegates core artboard syntax (repeats, branches, <sc-for>/<sc-if>/<dc-import>) to files that do not exist.

Trim Sharing and History to what the agent must do (share/comment commands, role semantics, the plain-http security caveat) and move the editor-UI narrative (merge-conflict bars, "Put my change back" flows, presence avatars) into a reference file.

Break the single giant canvas.json bullet in "The canvas's files" into labeled sub-bullets (boards entry, order, notes, pages/launch, designSystems) so the index rules are scannable rather than one run-on paragraph.

DimensionReasoningScore

Conciseness

The body is aggressively compressed and assumes competence throughout ("Rules that bite (each fails silently)", telegraphic rule lists, no explanation of concepts Claude already knows). It is not a 5 because the Sharing and History sections narrate editor UI behavior in detail the agent rarely needs to act on (merge bars, "Yours was kept", "Put my change back" flows, avatar cards), a noticeable amount of over-included material.

4 / 5

Actionability

Guidance is fully executable: exact CLI calls with their JSON outputs ("upload <root> <file> → {url:\"/_blob/<id>\", …}"), the exact canvas.json shape, a complete copy-paste-ready artboard HTML example, and a ready-to-write git pre-commit hook script. This matches the top anchor of copy-paste-ready commands covering the common cases.

5 / 5

Workflow Clarity

Creating is a numbered sequence (steps 1–6) with an explicit validation feedback loop — "run `check` and fix what it names until `problems` is 0" — and Revising repeats the pattern (steps 1–4, check until problems is 0). Error-recovery paths are also specified (failed uploads become labelled placeholders, NETWORK_BLOCKED handling), matching the anchor with explicit validation steps and feedback loops.

5 / 5

Progressive Disclosure

References are clearly signaled and one level deep with when-to-read guidance ("format.md (before a first artboard)", "diagrams.md (FIRST for diagrams)"), but none of the ten referenced artifact-type/reference/*.md files exist in the bundle — the only files there are google-fonts JSONs — so the critical rules delegated to them (repeats, branches, child components, craft) are unreachable. Signaling alone cannot rate 4 when the disclosed navigation is broken against the actual bundle structure, though the in-body organization keeps it above the minimal-structure anchors.

3 / 5

Total

17

/

20

Passed

Description

92%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong description: concrete multi-action capability statement, an explicit Use-when clause with format-specific triggers, and a clear negative boundary. The only room for improvement is slightly broader synonym coverage in the trigger phrasing.

DimensionReasoningScore

Specificity

The description lists multiple concrete actions with exact artifacts: "Make and revise designs in the Design (canvas) artifact format (project/canvas.json plus one .dc.html Design Component file per artboard) as local files, then open them in a local canvas editor where the user can view the artboards side by side, retype copy and adjust tweaks." Create, revise, open, and in-editor editing are all covered with no gaps, which matches the comprehensive-coverage anchor rather than the minor-gaps anchor at 4.

5 / 5

Completeness

It explicitly answers both questions: the what is the make/revise/open workflow above, and the when is a concrete "Use when the user asks for an editable canvas of artboards in this format, names claude-design or the Design (canvas) type, or wants to revise a folder that holds project/canvas.json." It even adds a negative boundary ("Not for a one-off HTML prototype..."), matching the top anchor exactly.

5 / 5

Trigger Term Quality

Triggers include "an editable canvas of artboards in this format", "names claude-design or the Design (canvas) type", and "revise a folder that holds project/canvas.json" — good natural-phrase coverage anchored on the skill's own name and format. It falls short of the 5 anchor because a few natural user phrasings (e.g., "design canvas", "design boards", "mockup folder") are not covered, but it is clearly above the missing-synonyms anchor at 3.

4 / 5

Distinctiveness Conflict Risk

The trigger is pinned to a specific file format (canvas.json plus .dc.html files) and explicitly excludes adjacent cases ("Not for a one-off HTML prototype or a design exploration outside this format"), giving it a clear niche with minimal conflict risk against generic design or HTML skills — the 5 anchor, not merely "mostly distinct".

5 / 5

Total

19

/

20

Passed

Validation

100%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository

JimLiu/baoyu-design

Path

chatgpt-plugins/baoyu-canvas/skills/claude-design/SKILL.md

Reviewed

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.