Loading ToolBench
MCP server for Cloudflare Flux Schnell worker API that implements a text-to-image generation tool using the Flux Schnell model
Single tool with reasonable naming and basic schema, but significant gaps in descriptions, error handling, and parameter documentation. The tool name 'generate_image' follows verb_noun convention appropriately. The input schema is present with JSON Schema types and constraints (minLength, maxLength). However, the tool description is generic and lacks actionable context (WHEN to use it, what to expect). Parameter description for 'prompt' is adequate but minimal. The server implements basic input validation via Zod but returns error messages as plain text rather than structured guidance. No output schema is documented. No error categorization or recovery guidance. The tool handles environment variable injection correctly (credentials not exposed as parameters), but overall structure is below production baseline (median 50-55 for community servers).
Generate an image from a text prompt using Flux Schnell model
Tool description lacks actionable context. Does not explain WHEN to call this tool, what happens to generated images, or dependencies (e.g., requires valid FLUX_API_URL/FLUX_API_TOKEN environment variables set beforehand). LLMs cannot reliably select or use this tool without richer guidance.
Output schema not documented. Tool saves image to disk and returns a plain text confirmation string. LLMs do not know what fields to expect or whether downstream tools can reference the saved file. No structured response (e.g., {filename, filepath, size, format, status}).
Error handling returns plain text messages without categorization or recovery guidance. E.g., 'Configuration Error: FLUX_API_URL environment variable is not set' gives the LLM no way to recover. Should state: 'Configuration Error: FLUX_API_URL not set. This server requires administrator configuration. Contact support.' Or ideally, catch this at startup, not on every tool call.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-21 | F | 49 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 11 | - | v1 |
Parameter 'prompt' description is minimal (7 words, 48 chars). Baseline is 72 chars. Does not specify constraints on content, prohibited topics, best practices for image generation, or what happens if prompt is empty/invalid. Guidance like 'Describe visual elements, composition, style, and color. Avoid copyrighted names and violent content.' would help LLMs craft better prompts.
No idempotency guarantees documented. Each call with the same prompt generates a different image (API behavior). If the LLM retries due to a transient error, it gets a new image. Should document this behavior and consider adding an optional 'seed' parameter for reproducible generation if the API supports it.
Irreversible operation (image generation consumes API quota, writes to disk) with no dry-run or confirmation step. An LLM in a loop could exhaust API limits or fill disk before detecting the issue. Should offer a 'dry_run' parameter or a separate 'validate_prompt' tool.