Gemini image scraper with CLI and MCP server surfaces.
Two tools with basic schemas and descriptions, but significant gaps in parameter documentation, error handling guidance, and output schema clarity. generate_image has 6 parameters but only 2 have descriptions in the visible code. login lacks parameter descriptions entirely. Neither tool documents return structure or error recovery paths. Descriptions are present but generic (194-char baseline suggests room for improvement). No evidence of input validation, error categorization, or recovery guidance.
Generate Gemini images using the saved NanoBanana session. Requires a valid Google Gemini login.
Open a visible browser window so a human can log into Google Gemini and save the session.
Parameter descriptions missing or incomplete. generate_image has 6 parameters (output_dir, model, tool, headless, debug) with no descriptions visible in schema. login's 'force' parameter lacks description. LLMs cannot infer parameter meaning from names alone.
No output schema documented. generate_image and login return dicts but structure is not declared. LLMs cannot plan downstream calls or extract required fields without knowing response shape.
No error handling guidance. generate_image catches AuthRequiredError and generic Exception but returns unstructured error dicts. No recovery hints (e.g., 'Call login() first'). Agents cannot self-correct.
Credentials and session state exposed via parameters. generate_image accepts 'headless' and 'debug' flags that control browser behavior; login forces interactive browser. No evidence of secure session injection or token handling.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 43 | 2026-07-28+ | v2 |
No input validation or constraint documentation. Parameters like 'model' and 'tool' accept free-form strings with no enum or format constraints. No min/max for numeric params. LLMs will hallucinate invalid values.