What problem does it solve?
Many teams and agents struggle to continuously generate, integrate, and verify design-driven web pages at scale; this Skill automates the baton-passing loop that reads a task, generates a Stitch screen, retrieves assets, sanitizes and integrates HTML, and writes the next task to keep the loop running.
Core Features & Use Cases
- Baton-based orchestration: Read and write the .stitch/next-prompt.md baton to coordinate iterative work between agents or CI/CD.
- Stitch MCP integration: Discover Stitch tools, create or reuse projects, generate screens from text prompts, and persist project/screen metadata in .stitch/metadata.json.
- Asset retrieval and safety: Download HTML and screenshots, perform a required safety audit to detect and sanitize scripts, then move production-ready pages to site/public and update navigation and SITE.md.
- Optional visual verification: Use Chrome DevTools MCP when available to render pages, capture screenshots, and compare fidelity against Stitch-generated images.
- Use Cases: Autonomous site generation in agent chains, CI/CD-driven page creation, and iterative design-to-production workflows for product teams and designers.
Quick Start
Read .stitch/next-prompt.md, generate the page with Stitch MCP, integrate the sanitized HTML into site/public, update .stitch/metadata.json and .stitch/SITE.md, and write the next .stitch/next-prompt.md to continue the loop.