MCP server for generating images using Gemini Web UI with Chrome automation
Two tools with basic schemas and descriptions, but significant gaps in parameter documentation and output structure. setup_gemini_auth has no input parameters documented (empty schema), violating the requirement that every parameter needs a description. gemini_image_generate has two parameters with types and descriptions, but output is unstructured JSON strings rather than documented schemas. No error handling guidance, no recovery paths, and descriptions lack context about when/why to use each tool. Tool names follow verb_noun pattern (positive), but descriptions are minimal (10-20 chars baseline). No pagination, no field chaining, no idempotency guarantees.
Generate an image using Gemini Web UI
Open Chrome GUI to login to Gemini. Run this once to authenticate.
setup_gemini_auth has empty input schema (no parameters documented). Tool lacks description of what happens (opens browser, waits for login) and when to call it (before first image generation).
Output schemas are unstructured JSON strings (e.g., '{"status": "success", ...}') rather than documented structured objects. LLMs cannot plan downstream calls or extract fields reliably. No documented return type.
Error responses are generic ('failed to generate image: %v') without recovery guidance. LLM cannot determine if error is retryable, user-fixable, or fatal. No actionable next steps provided.
gemini_image_generate description (27 chars) is below baseline (194 chars avg for A+ tools). Lacks context: when to call it, what prerequisites exist (must run setup_gemini_auth first), what the output contains, expected latency.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 46 | 2026-07-28+ | v2 |
No idempotency guarantees documented. If image generation partially fails or times out, agent cannot safely retry without risking duplicate images. No dry-run or confirmation pattern for destructive operations.