MCP Server for video analysis using Google Gemini API
Single tool 'analyze_video' has a reasonable description (Portuguese, ~280 chars) but lacks critical schema documentation. Input parameters are defined with types and descriptions in the code, but no explicit output schema is documented. The tool performs a destructive/write operation (creates .md files) but lacks error handling guidance, recovery instructions, or confirmation patterns. Parameter descriptions are adequate but lack format constraints or validation rules. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) present.
Analisa um arquivo de video usando a API Google Gemini. Faz upload do video, processa com o modelo Gemini e retorna uma analise estruturada contendo transcricao, descricao visual e resumo funcional. A transcricao tambem e gravada como um arquivo .md com o mesmo nome do video, na mesma pasta em que o video esta.
No output schema documented. Tool returns a string but LLMs cannot infer structure (transcription, visual description, functional summary). Downstream tools cannot chain on this output.
Destructive operation (writes .md files) lacks confirmation or dry-run pattern. No error handling guidance for file I/O failures, API errors, or invalid video formats.
Parameter 'instruction' lacks format constraints or examples. Description says 'optional' but does not explain what kinds of instructions are valid or how they affect analysis.
No tool annotations (destructiveHint, idempotentHint). LLMs cannot determine if repeated calls with same input are safe or if the tool modifies state.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 58 | <=2025-11-25 | v2 |
File path parameter accepts relative paths (e.g., 'modulo/aula.mp4') but no validation or error message for missing files, invalid paths, or path traversal attempts.