Server has 4 tools with complete Zod schemas and descriptions. Naming follows verb_noun pattern (getImageDescription*, getPhotoshop*, getPhotograph*). However, descriptions are verbose (200-300 chars) and lack actionable guidance on when to use each tool vs. alternatives. Parameter descriptions are present but generic. No output schema documentation. Error handling returns per-item success/failure but lacks recovery guidance. Missing tool annotations (readOnlyHint, idempotentHint). No pagination support despite accepting arrays. Baseline: 194 chars avg description (these are 2x that); 72 chars avg param annotation (these are ~150 chars, overly verbose).
Analyze image files for comprehensive visual analysis. Each path should point to an accessible image file that needs description. The analysis will examine visual elements, objects, colors, text, spatial relationships, context, and overall mood to provide an objective, detailed description suitable for accessibility purposes or content documentation.
Analyze images with custom instructions for analyzing the images. This prompt will determine what aspects of the image to focus on, what details to extract, and how the description should be structured. For example, you could ask for analysis of color palette, composition, technical aspects, emotional impact, or specific elements within the image.
Analyze photographs (JPG, PNG, RAW, etc.) for comprehensive photographic analysis. Each path should point to an accessible image file that needs technical and artistic evaluation for composition, technique, storytelling, and creative intent purposes.
Analyze Photoshop images (PNG, JPG) or individual layer exports (PNG) for design analysis. Each path should point to an accessible file that represents either a complete composition or isolated design element to be analyzed for layer composition, color schemes, and design hierarchy purposes.
Tool descriptions lack differentiation guidance. All four tools analyze images but descriptions do not explain when to use getImageDescription vs getPhotoshopImageDescription vs getPhotographDescription. LLMs will struggle to select the right tool.
No output schema documented. Tools return JSON with 'description', 'path', 'success', 'error' fields, but this structure is not declared in tool definitions. LLMs cannot plan downstream operations or validate responses.
Missing tool annotations. All tools are read-only (no state modification) but lack readOnlyHint annotation. This prevents agents from understanding safety properties and optimizing retry logic.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 54 | 2026-07-28+ | v2 |
Parameter descriptions are overly verbose (150+ chars) and repetitive. Baseline is 72 chars. Descriptions should be concise and actionable, not restate the tool description.
Error handling returns per-item success/failure but does not guide recovery. Error messages are generic ('Unknown error occurred'). Should include actionable guidance: 'File not found. Check path is absolute and file exists.'