Loading ToolBench
MCP server for Google Gemini image generation
Two tools with basic schemas and descriptions, but significant gaps in parameter documentation, output schema clarity, and error guidance. generate_images has a hardcoded Windows path in the default parameter and schema, creating portability issues. Neither tool documents output structure for LLM consumption. Descriptions are present but minimal (under 100 chars). No parameter descriptions for most inputs. Error handling exists but lacks recovery guidance.
Create HTML img tags from image file paths
Generate images using Google Gemini AI
Hardcoded Windows path in tool schema and implementation (G:\image-gen3-google-mcp-server\images). This breaks portability across operating systems and deployment environments. Default parameters must not contain environment-specific paths.
Output schemas not documented. generate_images returns {message, files} and create_image_html returns {html, message}, but LLMs cannot infer field types or structure. Downstream tools cannot reliably chain on these outputs.
Parameter descriptions missing or incomplete. 'imagePaths' in create_image_html has a description, but 'width' and 'height' lack context (units, constraints, typical values). 'numberOfImages' in generate_images lacks explanation of why 1-4 is the limit.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 40 | 2026-07-28+ | v2 |
Error messages lack recovery guidance. McpError(2, 'Failed to generate images: ...') tells the LLM nothing about what to try next. Should suggest: 'Check API key, verify prompt is not blocked, or try a simpler prompt.'
No input validation or constraint documentation. generate_images accepts any string prompt with no length limit or content policy guidance. create_image_html accepts any array of paths without validating file existence or format.