Primary question: Does your development environment utilize the Stitch MCP server or Agent Skills standard?
RepoDaily adoption score
RepoDaily rates this as 84/100 (strong) for adoption: evidence, installation path, production risk, differentiation, license clarity, and AI/agent fit are scored from the article sources and adoption notes.
4 source(s) across 2 source category/categories, plus a RepoDaily-specific evidence module when available.
3 workflow step(s), 3 next-action step(s), and 3 command/install signal(s) were detected.
Trending momentum is +338 stars, with maintenance/release/issue signals counted when present.
Risk is marked medium, with 3 security note(s) and 3 explicit skip condition(s).
2 opportunity lens item(s), 2 alternative(s), and 2 type-specific section(s) support differentiation.
License source or license wording is present.
7 AI/agent-related signal(s) were detected in the article text and metadata.
Project overview
Stitch Skills is a repository containing a collection of agent skills and plugins specifically designed for Google Stitch. The library is implemented in TypeScript and adheres to the Agent Skills open standard, which facilitates compatibility with a wide range of coding agents, including Codex, Antigravity, Gemini CLI, Claude Code, and Cursor.
The project provides pre-built capabilities for design and build workflows. It is structured to allow developers to install plugins globally or selectively integrate specific skills into their current projects. The skills are designed to automate tasks such as converting frontend code into Stitch designs, generating new design screens from text or images, and managing design systems through specific commands.
To function correctly, these skills require the Stitch MCP server to be configured and running within the agent's environment. The repository organizes skills into plugins like `stitch-design`, `stitch-build`, and `stitch-utilities`, each targeting different aspects of the development lifecycle. While hosted under the Google Labs Code organization, the project is explicitly documented as not being an officially supported Google product.
Why it is trending now
- Gained 338 stars recently, ranking 7th on the trend list, indicating strong interest in AI-driven development tools.
- Mentions compatibility with popular coding agents like Claude Code and Cursor aligns with the growing trend of AI-assisted software engineering.
- Provides a structured library for 'Agent Skills,' addressing the need for standardization in AI agent capabilities.
Problem it solves
- Developers often lack standardized, interoperable 'skills' that can work across different coding agents and environments.
- Converting existing frontend code into design formats or managing design systems within AI workflows is often manual and fragmented.
- Setting up AI agents to interact with specific tools like the Stitch MCP server typically requires bespoke configuration.
How it works
- Users install the plugin suite either globally via the `codex` CLI or locally for specific agents like Claude Code and Cursor using `npx` commands.
- The `codex` installation method supports sparse checkouts to limit downloads to specific paths like `.agents/plugins` or `plugins/stitch-design`.
- Once installed and the Stitch MCP server is running, users interact with skills using natural language prompts, such as asking to 'upload frontend code' or 'generate design variants'.
Installation and Setup Evidence
The repository offers two primary installation methods depending on the user's environment. For Codex users, the recommended approach involves adding the Stitch Skills marketplace via the CLI using the command `codex plugin marketplace add google-labs-code/stitch-skills --ref main`. This command supports optional `--sparse` flags to optimize cloning speed by checking out only necessary directories.
For users of Claude Code and Cursor, installation is handled via `npx`. The command `npx plugins add google-labs-code/stitch-skills --scope project --target claude-code` installs the plugins for the current project in Claude Code, while changing the target to `cursor` sets it up for a Cursor workspace. Selective installation of individual skills is possible using `npx skills add`, though the documentation notes that skills often have inter-dependencies that must be resolved.
Available Skills and Plugins
- stitch-design: Focuses on core design workflows. It includes skills like `stitch::code-to-design` for converting React/Vue code to Stitch Designs, `stitch::generate-design` for creating screens from text/images, and `stitch::manage-design-system` for uploading `DESIGN.md` files.
- stitch-build: Intended for build and component skills (details provided in plugin description).
- stitch-utilities: Covers utility and helper skills (details provided in plugin description).
- Dependencies: Selective skill installation requires users to manually verify and include all required dependencies, as the library notes frequent inter-dependencies between skills.
Who should pay attention?
Good fit if
- Teams already utilizing the Google Stitch platform or Stitch MCP server in their development pipeline.
- Developers using Claude Code, Cursor, or Codex who want to extend agent capabilities with standardized design and build skills.
- Projects requiring automated conversion of frontend code into design formats or automated design system management.
Skip for now if
- Organizations that do not use the Stitch MCP server or compatible coding agents.
- Teams seeking a standalone security tool, as this library acts as a plugin/skill collection for an existing ecosystem.
- Projects where inter-dependency management between agent skills is a critical bottleneck and cannot be manually handled.
Risks and cautions
The project requires a specific runtime environment (Stitch MCP) and is in an early version phase (0.1.x).
- Supported versions are currently limited to 0.1.x according to the security policy.
- The documentation explicitly states this is not an officially supported Google product, which may impact long-term support guarantees.
- Setup requires configuring the Stitch MCP server and handling potential inter-dependencies when installing skills selectively.
- Vulnerability reporting must be done through the GitHub Security Advisory tab, not standard Google channels.
- The project is not eligible for the Google Open Source Software Vulnerability Rewards Program (VRP).
- Supported version is 0.1.x; versions below 0.1 are not supported.
Alternatives to compare
| Approach | When to use | Trade-off |
|---|---|---|
Native Agent Plugins | When users prefer built-in tools specific to Claude Code or Cursor without MCP dependencies. | Varies |
Custom Skill Development | When teams have highly specific workflows that cannot be covered by the standard library. | High (Development Time) |
Direct Stitch MCP Workflows | When teams want to call Stitch MCP tools directly and keep prompts, dependencies, and workflow logic under their own control instead of adopting the plugin abstraction. | Lower dependency overhead, higher workflow maintenance |
What this trend reveals
Standardization of AI Interactions
The library's adherence to the 'Agent Skills' open standard suggests a move towards interoperable AI tools, reducing vendor lock-in for specific coding agents.
The README explicitly mentions compatibility with multiple agents (Codex, Antigravity, Gemini CLI, Claude Code, Cursor) via the Agent Skills standard.
Design-Dev Automation
Capabilities like `stitch::code-to-design` bridge the gap between frontend implementation and design tools, potentially streamlining the handoff process.
The `stitch-design` plugin documentation lists specific skills for converting code to design and extracting design systems from source code.
RepoDaily verdict
Stitch Skills offers a compelling, standardized library for extending coding agents with design and build capabilities, specifically tailored for the Stitch ecosystem. While the early version status (0.1.x) and lack of official Google product support present adoption risks, the open standard approach and compatibility with popular agents like Claude Code make it a valuable tool for teams investing in AI-assisted design workflows.