Loading ToolBench
AI-powered media operations server using Google Gemini multimodal capabilities. Specifically designed for coding agents and AI development workflows. Provides tools for generating creative content descriptions, analyzing images/videos/audio files, and creating detailed manipulation instructions. Enables coding agents to understand media content, generate media-related code, create comprehensive documentation, and provide intelligent media processing guidance. Supports advanced prompt engineering with structured outputs optimized for programmatic consumption.
Three tools with detailed descriptions but significant schema and naming issues. generate_media and analyze_media have verbose, marketing-focused descriptions (500+ chars) that waste tokens and bury actionable details. Schemas are present but lack critical constraints: no enums for format validation, no min/max bounds, no parameter descriptions in schema objects themselves. Tool names lack clear verb-noun structure (generate_media vs create_image, analyze_media vs describe_media). No output schemas documented. Error handling absent, no guidance on retryability, user-fixable vs fatal errors, or recovery paths. Composition is reasonable (three distinct concerns) but parameter naming inconsistencies (filePath vs outputFile) and missing chaining IDs in responses will force extra lookup calls.
Analyze images, videos, or audio files using Gemini's multimodal AI capabilities. Essential for coding agents building media-aware applications. CODING AGENT BENEFITS: - Generate structured metadata for content management systems and databases - Extract actionable insights for automated testing of UI/UX components - Create comprehensive image descriptions for accessibility compliance (WCAG) - Generate OCR data for document processing workflows and form automation - Analyze user-uploaded content for moderation and categorization systems - Extract technical specifications for media processing pipelines - Generate training data descriptions for machine learning model development - Create detailed API responses for media analysis services BEST PRACTICES FOR AI AGENTS - IMAGE ANALYSIS: - Object Detection: "List all objects visible in this image with their approximate positions (top-left, center, etc.)" - Scene Understanding: "Describe the setting, lighting conditions, time of day, and overall atmosphere of this image" - Composition Analysis: "Analyze the composition using rule of thirds, leading lines, symmetry, and focal points" - Color Analysis: "Describe the color palette, dominant colors, color harmony, and mood created by colors" - Quality Assessment: "Evaluate image quality: sharpness, exposure, white balance, noise, and areas for improvement" - Text Extraction (OCR): "Extract all visible text from this image, maintaining layout and formatting" - Face & Emotion Detection: "Describe visible people, their expressions, emotions, body language, and interactions" - Brand/Logo Recognition: "Identify any brands, logos, or trademarks visible in this image" - Accessibility Description: "Create a detailed alt-text description for visually impaired users" BEST PRACTICES FOR AI AGENTS - VIDEO ANALYSIS: - Content Summary: "Summarize the video content, key scenes, and narrative progression" - Scene Detection: "Break down the video into distinct scenes with timestamps and descriptions" - Action Recognition: "Describe all actions, movements, and activities occurring in the video" - Audio Description: "Describe what you hear: dialogue, music, sound effects, ambient sounds" - Production Analysis: "Analyze cinematography: camera movements, shot types, transitions, lighting, editing pace" - Accessibility Captions: "Generate detailed captions including speaker identification and audio cues" - Quality Check: "Identify technical issues: blur, shaky footage, poor lighting, audio problems" SUPPORTED FORMATS: - Images: JPEG, PNG, GIF, WebP (photos, screenshots, diagrams, artwork) - Videos: MP4, MPEG, MOV, AVI, WebM (up to ~10MB recommended) - Audio: MP3, WAV, AAC (speech, music, ambient sounds) PROMPT ENGINEERING TIPS: - Ask specific questions rather than "what's in this image?" - Request structured output: "List in bullet points" or "Create a table" - Specify detail level: "Brief overview" vs "Detailed frame-by-frame analysis" - Combine multiple aspects: "Analyze composition AND suggest improvements" - Request actionable insights: "What changes would improve this image?"
Descriptions are 500+ characters with marketing copy and best-practices guidance that should be in docs, not tool descriptions. Rubric baseline is 194 chars; these waste tokens and bury the core action. Example: generate_media description is ~1200 chars with 'CODING AGENT BENEFITS' and 'BEST PRACTICES' sections.
No output schemas documented. LLMs cannot plan downstream calls or extract required fields. generate_media returns 'absolute file path' and 'mime type' but schema is missing. analyze_media returns unstructured text. manipulate_media returns 'instructions' but format is undefined.
Parameter descriptions missing from JSON Schema objects. The 'prompt' parameter in generate_media has a description in the schema, but 'outputFile' lacks guidance on naming conventions, path handling, or extension behavior. Schema descriptions are present but inconsistent across tools.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 43 | 2026-07-28+ | v2 |
Generate actual images using Gemini's multimodal AI image generation capabilities. Optimized for coding agents building media-rich applications. CODING AGENT BENEFITS: - Generate actual images for UI/UX mockups and design prototyping - Create visual assets for automated content generation workflows - Produce product photography and marketing materials programmatically - Generate illustrations and graphics for documentation and presentations - Create visual training data for machine learning pipelines - Generate placeholder images and assets for development and testing BEST PRACTICES FOR AI AGENTS: - Image Generation: Be highly specific about style, composition, lighting, colors, mood, and subject details Example: "A photorealistic sunset over a mountain lake, golden hour lighting, reflection on calm water, 4K quality, dramatic sky" - Photorealistic Images: Use photography terms like camera angles, lens types (35mm, macro), lighting setups - Artistic Styles: Specify artistic styles (watercolor, sketch, digital art, impressionism) for non-photorealistic results - Product Photography: Describe studio setup, lighting, background, angle for product shots - Illustrations: Be specific about line style, color palette, shading technique - Technical Details: Mention desired quality level, level of detail, rendering style - Composition: Specify framing, rule of thirds, focal points, depth of field - Mood & Atmosphere: Describe the emotional tone and atmosphere you want to convey IMAGE OUTPUT: - Returns the absolute file path to the generated PNG/JPEG image - Images are saved to /tmp/gemini_mcp by default - File includes mime type information (image/png, image/jpeg, etc.) - Supports various aspect ratios and sizes
Provide intelligent instructions for transforming, editing, or enhancing images and videos using AI analysis. Perfect for coding agents building media processing workflows. NOTE: This tool generates INSTRUCTIONS for manipulation, not the actual edited media. Output is actionable guidance for image/video editing
No input validation constraints. Schemas lack enums for supported formats (JPEG, PNG, GIF, WebP for images; MP4, MOV, AVI, WebM for video). No min/max for file sizes, prompt length, or output filename length. LLMs will pass invalid formats and oversized files.
No error handling guidance. Tools do not document retryability, user-fixable errors (invalid API key, quota exceeded), or fatal errors (unsupported format). No recovery paths. LLMs cannot distinguish 'retry this' from 'ask the user' from 'give up'.