VibeGraphics MCP has 6 tools with basic schemas and descriptions present, but significant gaps prevent higher scoring. All tools have names starting with action verbs (banana_, veo_, vibe_), which is positive. However, parameter descriptions are present but often lack critical constraint information (ranges, enums, formats). Output schemas are not documented, responses are inferred from implementation rather than explicitly specified. Error handling is minimal. The source code shows tool registration via @mcp.tool() but the implementation is truncated, preventing full validation of error handling and output structure. Most tools accept file paths and model parameters but lack validation guidance. Tools are reasonably focused (one primary action each), but composition dependencies (vibe_plan_infographic → vibe_render_image_from_spec) are not explicitly documented as chains.

• Add explicit output schema documentation to each tool. Example for banana_generate: 'Returns object: {success: bool, image_paths: [str], file_sizes: [int], generation_time_seconds: float, error?: str}.'
• Document aspect_ratio and resolution enums for veo_generate_video. Example: 'aspect_ratio: one of ["16:9", "9:16", "1:1", "4:3"]. resolution: one of ["720p", "1080p", "1440p"].'
• Add min/max constraints to numeric parameters. Example: 'n: integer, 1 - 10 (best-effort; Gemini may return fewer); max_wait_seconds: 60 - 3600; max_code_files: 1 - 50.'
• Document the vibe_plan_infographic output format as a prerequisite for vibe_render_image_from_spec and vibe_animate_from_spec. Include a sample JSON structure or link to a schema.
• Add error recovery guidance. Example: 'If Gemini API fails due to quota, wait 60 seconds and retry. If GitHub returns 404, check that repo_url is public and branch exists.'
• Clarify prerequisites in tool descriptions. Example: 'Requires GEMINI_API_KEY environment variable set. Requires network access to api.github.com and ai.google.dev.'
• Add per-tool permission/scope declarations. Example: 'banana_generate requires Google Generative AI API access with image generation model enabled.'
• For vibe_fetch_github, document response structure: 'Returns JSON file path containing {owner, repo, branch, readme_text, code_files: [{path, language, content}]}.'
• Add idempotency guidance. Clarify whether repeated calls with identical inputs produce identical outputs or new generations. Critical for agent retry logic.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry