In the Skill's Own Terms
A brand-first landing page designer that runs a three-part interview (product purpose, brand feel, visual preferences), translates answers into a Stitch design system (mapping adjectives to colorVariant, fonts to headlineFont/bodyFont, shapes to roundness), generates desktop and mobile screens, leads iterative refinement through structured feedback (variants, edits, rollback), and delivers a deployment-ready bundle with versioned HTML, DESIGN.md, color tokens, and DEPLOY.md.
For users building single-purpose landing pages or product marketing sites who have no established visual direction. Not for dashboards, multi-page apps, component-level work, or users with existing design mockups.
What it produces
- Deployment-ready zip bundle containing index.html (desktop), mobile.html (if generated), DESIGN.md, color-tokens.json, user assets, and DEPLOY.md checklist
- DESIGN.md documenting the design system with brand feel, color direction with hex and rationale, typography choices, and shape roundness description
- color-tokens.json with structured design tokens: primary color, colorMode, colorVariant, fonts, roundness
- DEPLOY.md deployment checklist generated from the template in references/state-and-pitfalls.md
How It Works
- 01Verify Stitch SDK and Environment
Before starting the interview, the skill installs the SDK if missing, verifies the API key environment variable is set, and makes one minimal SDK call to confirm auth.
- 02Run three-part Brand Interview
Phase A captures product name, what it does, target users, and desired CTA. Phase B captures 3 brand adjectives and light/dark preference.
- 03Translate Interview to Stitch Parameters
The skill maps brand adjectives to colorVariant enum, light/dark to colorMode, font preference to headlineFont and bodyFont, and sharp/rounded to roundness enum using a translation table.
- 04Generate Screen and Open in Browser
The skill calls generate_screen_from_text with deviceType DESKTOP, saves the returned HTML as desktop-v1.html in .stitch/designs/, and opens it locally using the OS-specific command.
- 05Iterate Through Feedback Loop
After each generation, the skill asks three feedback questions, translates user feedback into edit_screens calls for targeted changes or generate_variants for exploration, and saves new versions with incremented filenames.
- 06Deliver Deployment Bundle
The skill creates a zip with the final HTML renamed to index.html, mobile.html if generated, DESIGN.md documenting the brand system, color-tokens.json, user assets, and a DEPLOY.md checklist.