Victor Design System logo

Victor Design System

Community
victorzhang016-code
victor-design-system

证据驱动的跨载体视觉设计与交付系统。用于海报、图文/社交内容、PPT/演示和产品 UI:理解主题与情绪,学习优秀人工参考,选择有依据且足够丰富的设计手法,制作 HTML 母版并完成可编辑交付与审查。

Overview

Publishervictorzhang016-code
Repositoryvictor-design
Skill namevictor-design-system
Stars
69
Forks
8
Bundled files
77
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 77 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by victorzhang016-code on GitHub. Read the source before you install it.

Installation

Install the Victor Design System AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/victorzhang016-code/victor-design/tree/main
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/victorzhang016-code/victor-design.git \
  .claude/skills/victor-design-system
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Victor Design System in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Victor Design System on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Victor Design System is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Victor Design

First principle — the richness chain

Understand the person, situation, tension, material world, and irreplaceable soul of the subject before choosing a visual direction. Then, in order:

  1. Subject and emotion first. Let what is true about the subject decide composition, typography, color, imagery, material, and production method.
  2. Every technique has a cause — subject cause or benchmark cause — and can state its viewer effect. "It looks designed" is not a cause.
  3. Enough techniques, not few. Richness and justification are co-equal requirements. A draft that passes every veto but feels thin has failed; do not answer AI slop with austerity.
  4. Reference-level before user-visible. Compare against strong human-made work of an adjacent family; a shown draft is already a designed artifact.

references/workflow/density-and-care.md owns the full method: what to add when the brief is thin, the benchmark obligation, and the density targets. Never begin from a style label, asset recipe, template, or AI spectacle.

Form sanity backstop — run before every adapter

Never let a model-written concept declaration decide its own deliverable form. Treat it as a hypothesis that must lose to user language, a task controller, platform convention, and ordinary reader sense.

  1. Read the user request and any controlling RUN.md, round control, or acceptance criteria. Quote the source that fixes the form, count, and editable delivery requirement.
  2. State the intended reader action in plain language.
  3. Counterfactual: if this carrier were the only thing the reader received, could it perform that action without the author explaining it? If not, reclassify before designing.
  4. Fallbacks when underspecified: social graphic-text/article → cover plus body pages; product task → related states including feedback; methodology/case/tutorial/pitch → narrated deck; poster/key visual → one canvas only when explicitly requested.
  5. Stop if the proposed form conflicts with a controller.

Task-grounding backstop

Before any new visual decision, establish a concise task brief per references/workflow/task-brief.md (TASK_BRIEF.md for substantial or unattended work). The brief fixes the authority for scope and facts, reader action and viewing conditions, content that must remain true, allowed materials, asset necessity, and the task-specific visual mother object. A prior result, a style name, or an attractive generated image cannot fill a missing answer. An unattended runtime records uncertainty and takes the least-assumptive path; it never invents user approval, facts, assets, or visual preferences.

Image-role preflight — zero step when an image is attached

An uploaded image does not determine the task type. Before form classification or any production, load references/workflow/image-role-routing.md and ask whether the image is the base/hero, project evidence, supporting material, or a reference/benchmark. Only Image role: base plus poster/key visual activates the image-as-carrier branch in references/adapters/poster.md.

Evidence, not a preset

Read references/style-evidence.md for every task. It ships the author's distilled visual grammar as a default evidence base; an adopter with approved work replaces it as step one of onboarding. It is a record of judgment, never a palette, font pair, or layout template. Use only evidence that fits the confirmed task, and name any exception. The default voice — its color courage, scale force, and materiality — always yields to the user's stated aesthetic preference; a user who asks for quieter, softer, or different is the highest authority on voice.

Required execution chain

  1. Image attached → image-role preflight above. Then the form sanity and task-grounding backstops.
  2. Read references/aesthetic-core.md, references/style-evidence.md, references/copy-discipline.md, references/workflow/density-and-care.md.
  3. Classify the form, then read exactly one adapter: references/adapters/poster.md (user-specified single canvas), graphic-text.md (cover-plus-body editorial/social), product-ui.md (product surfaces), slides.md (cases, methods, tutorials, pitches).
  4. Read references/operations/execution.md and references/operations/three-gates.md — blocking rules, not paperwork to complete afterward.
  5. Conduct the design dialogue through the native structured prompt when callable; otherwise ask in grouped plain prose. Never infer Gate 1 approval from the request.

Conditional tools — load only for the named need:

NeedLoad
Asset sourcing, cutout, generation, SVG, perspectiveoperations/production-toolkit.md (after Gate 1 release)
Figma/PPTX/interactive-HTML build pipelinesoperations/delivery-implementations.md (at Gate 3)
Editable translation, font/material/perspective driftoperations/figma-fidelity.md
Any master review or completion claimoperations/review.md
References to decompose, or revision-led workoptional/reference-synthesis.md
Multi-format communication surfacesoptional/communication-surfaces.md
Long-running project governanceworkflow/project-governance.md

Precedence: user-confirmed subject and acceptance criteria → truth, safety, accessibility, primary task → approved project/surface decisions → this file, aesthetic core, style evidence → adapter and operations → optional references and external defaults.

Non-negotiable defaults

  • Asset source chain, in order: local workspace material → licensed web → user-approved AI generation → code (CSS/SVG/HTML, material/graphic roles only). Generation is never the default hero and never factual evidence.
  • A generated or found object is input material, not a poster. Process, crop, grade, mask, and seat it inside an authored canvas that still works with the object hidden.
  • Derive every palette and type voice from current-subject evidence. A prior success is not evidence for a new project unless reuse is explicitly justified and approved.
  • Title, required copy, source evidence, and refusal list are source of truth. No filler metadata or uncaused devices; factual inscriptions are carried as crafted small type.
  • Never batch-produce a multi-page deliverable before user-approved pilot pages. A page below the recorded density target is unfinished.
  • Use HTML to judge masters; obtain explicit render approval before any translation. Every category ships its native editable deliverable unless the user explicitly accepts flattened-only.
  • The approved render is Gate 3's golden source. Inventory custom fonts, masks, blends, and perspective composites before translation.
  • A control record proves a decision happened; it never replaces user approval or authored visual judgment.

Review order

Subject specificity, material relation, and residue first; hierarchy, type, color, space, crop, and task clarity second; technical checks last. Deterministic checks find concrete risks but cannot certify taste.

Architecture

This file is the runtime entry point. The default chain is aesthetic-core, style-evidence, density-and-care, one adapter, execution, and three-gates. Keep detailed rules in the referenced module that owns them; each rule lives in exactly one place.

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

and 17 more files.

Frequently asked questions

What does the Victor Design System AI skill do?

证据驱动的跨载体视觉设计与交付系统。用于海报、图文/社交内容、PPT/演示和产品 UI:理解主题与情绪,学习优秀人工参考,选择有依据且足够丰富的设计手法,制作 HTML 母版并完成可编辑交付与审查。

Why use Victor Design System on TypingMind?

Because you install it once and use it with any model. Victor Design System is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Victor Design System in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/victorzhang016-code/victor-design/tree/main. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Victor Design System?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Victor Design System?

As many as you like. As long as a model supports skills, you can use Victor Design System with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Victor Design System AI skill free?

Yes. It is published on GitHub by victorzhang016-code under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇