version: "1.0.0" name: gemini-file-analysis description: Analyze files (PDFs, images, audio, video, large source files) via Gemini's multi-modal file analysis. Use when a file is too large for Claude's context or when the file is a non-text format requiring visual/audio understanding.

Gemini File Analysis

Use this skill to send files to Gemini for multi-modal analysis. Gemini can read PDFs, interpret images, transcribe audio, understand video frames, and reason over large source files that exceed Claude's context window.

When to use this skill

• PDF documents: Research papers, contracts, architecture decision records, runbooks, changelogs.
• Images: Screenshots with error messages, diagrams, UI mockups, charts, or photos requiring visual interpretation.
• Audio files: Meeting recordings, user interviews, voiceovers that need transcription or content extraction.
• Video files: Screen recordings, demo clips, or short videos requiring frame-level understanding.
• Large source files: A codebase archive, minified bundle, or log file that is too large to fit in Claude's context.
• Binary formats: Any file format Claude cannot read as plain text.

MCP tools

Usage pattern

Analyze a PDF

{
  "tool": "gemini_analyze_file",
  "arguments": {
    "file_path": "/path/to/architecture-decision-record.pdf",
    "prompt": "Summarize the key architectural decisions and list any open questions."
  }
}

Interpret a screenshot

{
  "tool": "gemini_analyze_file",
  "arguments": {
    "file_path": "/path/to/error-screenshot.png",
    "prompt": "What error is shown and what are the most likely root causes?"
  }
}

Transcribe an audio recording

{
  "tool": "gemini_analyze_file",
  "arguments": {
    "file_path": "/path/to/meeting-recording.mp3",
    "prompt": "Transcribe the key action items and decisions from this meeting."
  }
}

Analyze a large source file

{
  "tool": "gemini_analyze_file",
  "arguments": {
    "file_path": "/path/to/large-codebase.tar.gz",
    "prompt": "Identify the main modules, entry points, and any obvious code quality issues."
  }
}

Model selection guidance

• For image and text files, the default model (gemini-3.1-pro-preview) is sufficient.
• For faster, cheaper answers on simple files, pass gemini-3.8-flash via the model parameter.
• Files over 15MB must use a gemini-2.5-* model. Above that size the tool falls back to Google's Files API, and Gemini 3.x models reject Files API references with 403 PERMISSION_DENIED. Smaller files are sent inline and work on every model.
• Files consume tokens proportional to size; avoid sending files unnecessarily.

Tips

• Be specific in the prompt. "Summarize this PDF" is less useful than "List all API endpoints documented in this PDF and their authentication requirements."
• For images containing text (screenshots, diagrams), ask Gemini to extract the text explicitly if you need to process it further.
• Audio transcription quality depends on recording quality; mention any domain-specific terminology in the prompt to improve accuracy.
• If the file path is a URL rather than a local path, check whether gemini_analyze_file supports URL inputs or whether you need to download the file first.

Do NOT use this skill for

• Files that fit comfortably within Claude's context window as plain text (just read them directly).
• Generating new images, video, or audio (use gemini-image-gen, gemini-video-gen, or gemini-audio-tts-music).
• Live web research (use gemini-research-grounded).
• Running code or doing math (use gemini-code-exec).
• Simple text-only questions where no file is involved (use gemini-chat-and-reason).