dev-design-context

name: dev-design-context description: 'Use as a one-time design setup for UI, product, landing-page, or brand-heavy work. Scans the project for design context, asks only the UX questions that cannot be inferred, then writes persistent design guidelines to .design-context.md. Does not implement features or review code.'

Dev Design Context

Gather design context for this project, then persist it so future UI work starts with the same visual and product direction.

This skill is for one-time setup or major design-direction resets. It does not write application code, generate mockups, or review implementation diffs.

Trigger routing

Use this skill when the user wants Codex / Claude Code to learn a project's design direction before building UI.

Trigger phrases include:

• dev-design-context
• setup design context
• 设计上下文
• 设计规范
• 先建立设计方向
• gather design context
• persistent design guidelines

Output goes to .design-context.md in the project root.

Step 0 — Load baseline

执行前先加载 references/dev-baseline.md。以下行为准则在本 skill 全程生效:不假设、最小代码、外科手术式改动、可验证成功标准。

baseline 与本 skill 的关联点:

• 不假设 —— 先扫描真实代码和资产,再问问题;不要让用户重复回答代码里已经能看到的信息。
• 最小代码 —— 只写 .design-context.md 的设计上下文,不要顺手改组件、CSS 或产品文案。
• 外科手术式改动 —— 如果 .design-context.md 已存在,只更新 ## Design Context 相关内容。
• 可验证成功标准 —— 结束前明确写出设计原则已经落在哪个文件,并总结未来设计工作应遵守的 3-5 条原则。

Step 1 — Explore the codebase

Before asking questions, scan the project to discover what can be inferred:

• README and docs: project purpose, target audience, and stated goals
• Package / config files: tech stack, dependencies, and design libraries
• Existing components: layout patterns, spacing, typography, component conventions
• Brand assets: logos, favicons, image assets, and already-defined color values
• Design tokens / CSS variables: color palettes, font stacks, spacing scales, radii, motion values
• Style guides or brand docs: any explicit product, content, or visual direction

Summarize what you learned and what remains unclear before asking the user anything.

Step 2 — Ask UX-focused questions

Ask the user directly to clarify only what cannot be inferred from the codebase. Keep questions focused and avoid long questionnaires.

Users and purpose

• Who uses this? What is their context when using it?
• What job are they trying to get done?
• What emotions should the interface evoke, such as confidence, calm, delight, urgency, or precision?

Brand and personality

• How would you describe the brand personality in 3 words?
• Are there reference sites or apps that capture the right feel? What specifically works about them?
• What should this explicitly not look like? Are there anti-references?

Aesthetic preferences

• Are there strong preferences for visual direction, such as minimal, bold, editorial, technical, playful, or organic?
• Should it support light mode, dark mode, or both?
• Are there colors that must be used or avoided?

Accessibility and inclusion

• Are there specific accessibility requirements, such as a WCAG level or known user needs?
• Are there constraints around reduced motion, color blindness, contrast, or other accommodations?

Skip questions where the answer is already clear from the codebase exploration.

Step 3 — Write design context

Synthesize the codebase findings and the user's answers into this section:

## Design Context

### Users
[Who they are, their context, and the job to be done]

### Brand Personality
[Voice, tone, 3-word personality, and emotional goals]

### Aesthetic Direction
[Visual tone, references, anti-references, theme, color, and motion direction]

### Design Principles
[3-5 principles derived from the codebase and conversation that should guide all design decisions]

Write this section to .design-context.md in the project root. If the file already exists, update the ## Design Context section in place instead of duplicating it.

Then ask the user whether they also want the same Design Context appended to .github/copilot-instructions.md. If yes, append or update that section there as well.

Confirm completion and summarize the key design principles that will guide future UI work.

Multi-Agent Profile

Recommended agent_type: explorer

Use when:

• The main agent needs design-system or UI convention discovery before UI work.
• The project has existing components, styles, assets, or brand cues to inspect.
• The sub-agent can summarize findings without implementing UI changes.

Do:

• Explore README, assets, styles, components, tokens, and existing screens.
• Separate codebase facts from assumptions.
• Ask only UX questions that cannot be inferred.
• Follow the explorer boundaries in ../../docs/multi-agent-policy.md.

Do not:

• Implement UI.
• Rewrite product copy or CSS outside .design-context.md.
• Invent brand direction when the project already provides evidence.

Output:

• Codebase design facts
• Unknowns that require user input
• Proposed design principles
• Files inspected