MCP server that audits the images on any web page — natural vs rendered dimensions, oversized-image detection, and format breakdown. Powered by imagedimensions.com.
Single tool with clear naming and good schema structure, but description lacks depth for production use. The tool name 'scan_image_dimensions' is well-formed (verb_noun), but description, while present, could better guide LLM decision-making. Parameters are typed and constrained, but lack detailed guidance. Output structure is partially documented through code inspection but not formally declared in the schema. Error handling is present but minimal. Overall a fair implementation typical of specialized single-function tools.
Audit the images on any public web page. Returns each image's natural vs rendered dimensions, flags oversized images (downloaded much larger than displayed — a common Largest Contentful Paint problem), and breaks down formats (WebP/AVIF adoption). Useful for web performance and image-optimization work.
Output schema not formally documented in tool registration. While code shows structured output (url, reportUrl, analysis fields), the tool schema lacks an explicit outputSchema declaration. LLMs cannot predict return structure and must infer from description.
Description does not explain when NOT to use this tool or prerequisite conditions. Lacks clarity on what 'public web page' means (auth-protected sites will fail silently). No mention of timeout behavior (120s limit visible in code but not exposed to LLM).
Error handling returns generic text without guidance. catch block wraps all errors in 'Image scan failed: {message}' but does not distinguish between timeout (retry), invalid URL (user error), or API service failure (contact support). No recovery guidance.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | 2025-06-18+ | v2 |
Parameter 'oversizedThreshold' description uses technical jargon ('Area-overshoot ratio', '≈2x in each dimension') that an LLM may not translate into valid calls. For a specialized tool this is acceptable, but guidance on typical values is missing.