stitch
Generate UI designs and frontend code with Google Stitch via MCP. Use when asked to create screens, mockups, UI designs, or generate frontend code from text descriptions. Supports desktop, mobile, and tablet layouts.
88
82%
Does it follow best practices?
Impact
100%
2.12xAverage score across 3 eval scenarios
Advisory
Suggest reviewing before use
Quality
Discovery
100%Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.
This is a strong skill description that clearly communicates what the skill does, when to use it, and through which tool. It uses natural trigger terms that users would actually say, specifies the tool (Google Stitch via MCP) for distinctiveness, and includes an explicit 'Use when' clause with good keyword coverage.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Lists multiple specific concrete actions: 'Generate UI designs', 'frontend code', supports 'desktop, mobile, and tablet layouts'. Names the tool (Google Stitch via MCP) and the types of outputs clearly. | 3 / 3 |
Completeness | Clearly answers both what ('Generate UI designs and frontend code with Google Stitch via MCP') and when ('Use when asked to create screens, mockups, UI designs, or generate frontend code from text descriptions') with explicit trigger guidance. | 3 / 3 |
Trigger Term Quality | Includes strong natural trigger terms users would say: 'screens', 'mockups', 'UI designs', 'frontend code', 'text descriptions', 'desktop', 'mobile', 'tablet'. These cover common variations of how users would phrase requests. | 3 / 3 |
Distinctiveness Conflict Risk | Clearly scoped to UI design generation via a specific tool (Google Stitch via MCP), which creates a distinct niche. The combination of UI design + code generation + specific MCP tool makes it unlikely to conflict with generic coding or design skills. | 3 / 3 |
Total | 12 / 12 Passed |
Implementation
64%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a solid, actionable skill with excellent concrete examples and copy-paste-ready commands covering the full Google Stitch MCP workflow. Its main weaknesses are moderate verbosity (some repeated information and sections that could be externalized) and a lack of explicit validation checkpoints in multi-step workflows. The prompt engineering and device type sections, while useful, contribute to document length that would benefit from progressive disclosure into separate files.
Suggestions
Add explicit validation steps between workflow stages (e.g., 'Verify generation completed by calling get_screen before fetching code') to create proper feedback loops
Move the prompt engineering guide, device types table, and error reference into separate linked files to reduce the main skill's token footprint
Remove the 'Weak Prompts' examples — Claude already understands what makes a vague prompt; the strong examples alone suffice
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is mostly efficient but includes some sections that could be tightened. The tools table is useful but the prompt engineering section with 'weak prompts' examples and the device types table add moderate bloat. The content is generally well-structured but could be more compact in places like the model selection section which repeats timing info already stated in the parameters section. | 2 / 3 |
Actionability | Every workflow includes fully executable, copy-paste-ready commands with specific parameter examples. The prerequisites section provides concrete setup steps, the examples section shows complete end-to-end flows, and parameter documentation is thorough with actual values and types. | 3 / 3 |
Workflow Clarity | The iteration workflow provides a clear sequence but lacks explicit validation checkpoints. The 'Complete Design Flow' example shows steps but doesn't include verification between steps (e.g., checking if generation succeeded before listing screens). The response handling section mentions suggestions but doesn't provide a clear decision tree. The note 'Do not retry' for generation is good but more feedback loops would strengthen this. | 2 / 3 |
Progressive Disclosure | The content is well-organized with clear headers and logical sections, but it's a fairly long monolithic document (~180 lines) that could benefit from splitting detailed reference material (tools table, device types, error table, prompt engineering guide) into separate files. No external references are used despite the content length warranting it. | 2 / 3 |
Total | 9 / 12 Passed |
Validation
90%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 10 / 11 Passed
Validation for skill structure
| Criteria | Description | Result |
|---|---|---|
metadata_version | 'metadata.version' is missing | Warning |
Total | 10 / 11 Passed | |
- Repository
- Dicklesworthstone/pi_agent_rust
- Commit
47823e3
- Reviewed
Table of Contents
Is this your skill?
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.