Single tool 'upload' has a complete input schema with proper types and descriptions for all parameters. Tool description is detailed (280+ chars) and explains the upload mechanism, output URL format, and streaming behavior for large files. However, the tool lacks output schema documentation, the response structure is not formally declared, forcing LLMs to infer field meanings from JSON. Parameter descriptions are adequate but could be more prescriptive about constraints (e.g., file size limits, supported formats). No error recovery guidance is provided. The tool performs a single, well-defined action (upload media), but lacks idempotency hints and confirmation patterns for a destructive operation.
Upload media (images/videos) to Cloudinary. For large files, the upload is processed in chunks and returns a streaming response. The uploaded asset will be available at: - HTTP: http://res.cloudinary.com/{cloud_name}/{resource_type}/upload/v1/{public_id}.{format} - HTTPS: https://res.cloudinary.com/{cloud_name}/{resource_type}/upload/v1/{public_id}.{format} where {cloud_name} is your Cloudinary cloud name, resource_type is 'image' or 'video', and format is determined by the file extension.
Output schema not documented. Response fields (public_id, version, signature, format, resource_type, created_at, bytes, type, url, secure_url) are returned but not formally declared in tool definition. LLMs cannot plan downstream operations or validate response structure.
No error recovery guidance. Errors are wrapped in McpError with generic messages like 'Upload failed: <error>'. LLM receives no actionable next steps (retry? ask user? check file format?). Should categorize errors as retryable vs. user-fixable.
Missing parameter constraints. 'file' parameter accepts 'Path to file, URL, or base64 data URI' but does not specify supported file types, size limits, or format restrictions. LLMs may pass invalid inputs without guidance.
No idempotency or confirmation pattern. Upload is a write operation that modifies state (creates asset in Cloudinary). No dry-run, confirmation step, or idempotency hint provided. Agents may accidentally retry and create duplicate uploads.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 55 | 2026-07-28+ | v2 |
Credentials exposed in environment variables without validation. Code checks CLOUDINARY_CLOUD_NAME, CLOUDINARY_API_KEY, CLOUDINARY_API_SECRET at startup but does not validate they are non-empty strings or handle missing credentials gracefully at runtime.