Benchmarks: Answer 99.16% of DocVQA Without Images in QA: Agentic Document ExtractionRead more

Build Faster with ADE Claude Skills

Teach your AI coding agent to build document workflows with ADE

Ava Xia

Ava Xia

Share On :
Build Faster with ADE Claude SkillsBuild Faster with ADE Claude Skills

TL;DR

  • A skill, not a wrapper. The ADE Claude Skill teaches Claude how to use LandingAI's Agentic Document Extraction (ADE) correctly — Parse, Build Schema, and Extract — so the code it writes follows best practices.
  • One-line install. Add the plugin marketplace, install the plugin, and reload. Claude picks up the skill automatically when you describe a document task.
  • Prompt library included. After installing, point Claude at ready-made prompts for the most common workflows: quickstart, batch-to-CSV, invoice processing, and a grounding viewer app.
  • Watch it work. A short video walkthrough shows the skill going from an empty folder to a working extraction pipeline.

Why a Skill?

Large language models are good at writing code, but they don't know the shape of your APIs. Ask a general-purpose agent to "extract fields from these PDFs" and it will guess — inventing endpoints, skipping schema design, or ignoring grounding entirely.

The ADE Claude Skill closes that gap. It's a set of instructions and references that teaches Claude Code how ADE actually works: when to Parse, how to Build a schema, how to Extract, and how to wire those calls into a real pipeline. Once installed, Claude recognizes document-processing tasks and reaches for ADE the right way — no hand-holding required.

Video Walkthrough

Here's the skill in action, from install to a working extraction pipeline:

Installing the Skill

If you use Claude Code, setup is three commands:

/plugin marketplace add landing-ai/ade-document-processing-skills
/plugin install ade-document-processing@ade-document-processing-skills
/reload-plugins

That's it. The skill activates on its own whenever you describe a document-processing task — parsing a PDF, extracting fields with a schema, classifying a batch, or visualizing grounding. For deeper reference, you can also connect the ADE MCP server for token-efficient access to the full documentation.

For the slides and the rest of the "build with AI agents" walkthrough, see the ADE docs.

Prompts to Try After Installing

Installing the skill gives Claude the knowledge. The prompt library in ade-sample-projects gives it a starting point. Copy any prompt into Claude Code, adjust the folder paths to match your project, and run:

  • Quickstart — the minimal three-call sequence (Parse → Build Schema → Extract) on a single document. Start here to learn the APIs.
  • Batch to CSV — scales the same flow across a folder of documents and writes one combined CSV using the Python SDK.
  • Invoice Processing — an async pipeline (Parse Jobs → Extract Jobs) that produces field, line-item, and metadata CSV summaries for large invoice batches.
  • Grounding Viewer App — a full-stack app with a side-by-side document and response viewer plus hover-to-highlight grounding, so you can visually verify every extracted value against the source page.

Because the skill is already installed, the code Claude generates from these prompts follows ADE best practices out of the box.

Getting Started

  1. Install the skill with the three commands above.
  2. Grab a prompt from the prompt library and drop it into Claude Code.
  3. Point it at your own documents and let Claude build the pipeline.

Read more in the ADE documentation.