# media-interpreter

media-interpreter — a subagent for working with PDFs, images, diagrams, and other binary formats. Extracts text, tabular data, and descriptions of visual content when standard file reading isn't an option.

- subagent · `type=agent` · Claude Code, Codex, Cursor · ★ 1064 repo stars
- Author: jarrodwatts
- License: not stated · policy=link-only · description only; the file stays with the author
- Source: https://github.com/jarrodwatts/claude-code-config/tree/HEAD/agents/media-interpreter.md
- Updated: 2026-01-21
- Security check: automated, 2026-11-16 (text) — no signs of malicious code found.
- HTML: https://affpapa.org/en/nevibecode/agents/media-interpreter

## What it does

media-interpreter runs when you need information from a file that Read cannot parse as plain text — PDFs, error screenshots, architecture diagrams, financial reports with tables. The subagent analyzes a file based on your specific goal and returns only the relevant data: extracted text, structured tables, descriptions of UI elements, or relationships between diagram components.

The subagent has access to web search, code context, and repository tools to verify or supplement extracted information when needed. It never exposes raw file contents to the main agent — only what you requested — to conserve context tokens.

Do not use media-interpreter for source code, plain text, markdown, or JSON; use Read instead. This subagent is for cases where Read would return garbled output or when you need selective data from a complex document.

## Install

### Claude Code (native platform)

**Manually: put the files in place**
- File: `.claude/agents/media-interpreter.md` — For all projects use ~/.claude/agents/. Claude delegates by the description field or an @-mention

### Codex

**Manually: put the files in place**
- File: `.codex/agents/media-interpreter.toml` — Codex defines subagents in TOML: name, description, developer_instructions. Move the Markdown text into developer_instructions by hand. For all projects use ~/.codex/agents/

### Cursor

**Manually: put the files in place**
- File: `.cursor/agents/media-interpreter.md` — Cursor also reads .claude/agents/

Where a file is given instead of a command, copy the files from the source to that path: https://github.com/jarrodwatts/claude-code-config/tree/HEAD/agents/media-interpreter.md

## When to use it
- pull text and tables from a PDF document
- read an error from a screenshot and describe its context
- analyze an architecture diagram and explain service connections
- extract structured data from a financial report or registry
- describe a UI mockup from an image

## Example prompts
- Which authentication methods are listed in `docs/api-spec.pdf`? Use media-interpreter.
- I have an error screenshot at `error.png`. Identify and describe what's happening.
- Explain how the services communicate based on `architecture.png`.

## Similar
- [datasheet-interpreter](https://affpapa.org/en/nevibecode/skills/datasheet-interpreter.md) — skill · 173 installs
- [social-media-image-sizes](https://affpapa.org/en/nevibecode/skills/social-media-image-sizes.md) — skill · 1900 installs
- [media-asset-management](https://affpapa.org/en/nevibecode/skills/media-asset-management.md) — skill · 302 installs
- [media-analysis](https://affpapa.org/en/nevibecode/skills/media-analysis.md) — skill · 127 installs
- [Contenta Photo Converter](https://affpapa.org/en/nevibecode/mcp/contenta-photo-converter.md) — MCP server
- [remove-ai-marks](https://affpapa.org/en/nevibecode/skills/remove-ai-marks.md) — skill · 1200 installs

From the same set: [jarrodwatts/claude-code-config](https://affpapa.org/en/nevibecode/by/jarrodwatts/claude-code-config.md) — 20 cards

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md
