Skip to main content
ClaudeWave
ES, Spanish
Skill1.4k estrellas del repoactualizado 6d ago

chat-complex-documents

Chat with and search your complex documents — ask questions, extract tables and fields, and get answers grounded in the source. Connects the hosted Unstructured Transform MCP server to parse, structure, and enrich PDFs, Word/Excel/PowerPoint, images, scanned files, emails, and 60+ other formats into clean, AI-ready text for search, Q&A, and summarization without building a custom parsing pipeline.

Instalar en Claude Code
Copiar
git clone --depth 1 https://github.com/vellum-ai/vellum-assistant /tmp/chat-complex-documents && cp -r /tmp/chat-complex-documents/skills/chat-complex-documents ~/.claude/skills/chat-complex-documents
Después abre una sesión nueva de Claude Code; el skill carga automáticamente.

SKILL.md

## What this does (in plain terms)

You have documents — contracts, reports, slide decks, scanned PDFs, spreadsheets, emails, images — and you want the assistant to reliably get the content _out_ of them and make it usable: searchable, answerable, summarizable, or ready to feed a knowledge base or another tool.

**Unstructured Transform** does exactly that. Hand it files and it turns them into clean, structured, AI-ready data across 60+ formats with one pipeline — no custom parsing or OCR to build and maintain. See the [Transform overview](https://docs.unstructured.io/transform/overview).

Under the hood it:

- **Partitions** each document into structured elements (titles, paragraphs, tables, lists), adjusting extraction per page for accuracy and cost.
- **Enriches** the result with metadata, table and image descriptions, and entity recognition.

The result is structured JSON that's ready for search, Q&A, summarization, and agents.

## Why people use it

- **Make documents searchable and answerable.** Transform extracts clean, structured text and tables from your files so the assistant can search them and answer questions grounded in the source.
- **Handle messy, varied files.** One integration covers 60+ formats (scanned PDFs, tables, images, Office files) instead of a per-format parser.

## When to use

USE THIS SKILL WHEN:

- The user wants to extract or structure content from documents (PDF, Office, images, emails, scanned files).
- The user wants documents made searchable or ready for RAG / Q&A / a knowledge base.
- The user asks to connect Unstructured or Unstructured Transform.
- A Transform tool returns an auth error → re-run the `auth` step below.

## Prerequisites

- An **Unstructured account** to sign in with during the OAuth step (a free tier is available). See [pricing](https://docs.unstructured.io/transform/billing).
- Nothing to install locally — Transform is a hosted MCP server.

## Setup

### Step 1 — Detect your environment

Determine which command tool to use for every command in this skill:

- If `host_bash` is available → you are on the **desktop app**. Use `host_bash` for all commands (the `auth` step opens your local browser).
- If it is unavailable → you are on the **web app**. Use `bash` for all commands (the platform handles the browser redirect).

### Step 2 — Add the server

Add Transform as a remote `streamable-http` MCP server. Keep the default risk level (`high`) — Transform ingests file content, so per-call approval is the safe default.

```
assistant mcp add unstructured-transform -t streamable-http -u https://mcp.transform.unstructured.io
```

The server is registered under the id `unstructured-transform` — that's the name shown by `assistant mcp list` and used in the commands below, even though this skill is titled "Chat with Complex Documents".

### Step 3 — Authenticate via OAuth

```
assistant mcp auth unstructured-transform
```

This opens Unstructured's authorization page. After sign-in, the assistant handles the callback and caches the tokens.

- On **desktop** → run via `host_bash` (opens your local browser).
- On **web app** → run via `bash` (the platform handles the redirect).

### Step 4 — Verify

```
assistant mcp list
```

Confirm `unstructured-transform` shows `✓ Connected`. If it shows `! Needs authentication`, re-run Step 3. If its tools don't appear within ~10 seconds, run `assistant mcp reload`.

## Using Transform

Once connected, ask the assistant in plain language, e.g. _"Use Unstructured Transform to parse these files and return one JSON file per source file,"_ or _"Extract the tables from this PDF."_ Transform partitions each file and returns structured output; to tune partitioning and enrichment behavior, see [Control Transform output](https://docs.unstructured.io/transform/output).

Per-request limits (Transform enforces these; plan around them):

- Each file must be a [supported file type](https://docs.unstructured.io/transform/supported-file-types).
- Each file must be **50 MB or less**.
- Each request may include **10 files or fewer**.
- At most **5 requests** may run at a time.

## Disconnecting

```
assistant mcp remove unstructured-transform
```

This removes the config entry and cleans up the stored OAuth credentials.

## SKILL COMPLETE WHEN

- **Connect:** `unstructured-transform` appears in `assistant mcp list` as `✓ Connected` and the user confirms its tools are available in the conversation.
- **Disconnect:** `assistant mcp remove unstructured-transform` succeeds and the server no longer appears in `assistant mcp list`.
acpSkill

Set up, authenticate, and run external coding agents (Claude Code, Codex) via the Agent Client Protocol

amazonSkill

Shop on Amazon and Amazon Fresh through your browser

api-mappingSkill

Record and analyze API surfaces of web services

app-builderSkill

Build and edit small, personal visual tools and artifacts — dashboards, trackers, calculators, data visualizations, charts, simple landing pages, and slide decks the user wants for THEMSELVES. This is the right skill whenever the user asks to "visualize this," "make a chart," or "build an artifact" for their own use, or to edit an app they already built here. Do NOT reach for a ui_show dynamic_page to fake an artifact — build a real persistent app here. A one-off diagram, chart, or explainer that just accompanies an answer belongs to the visualize skill (ui_show visual), not an app. NOT for complex, multi-user, or shippable products — those go to a real project folder with a coding agent (see Scope below).

app-controlSkill

Drive a specific named macOS app via raw input bypassing the Accessibility tree

assistant-migrationSkill

Migrate from ChatGPT, Claude, OpenClaw, Hermes, Manus, and other AI assistants into Vellum by inspecting their data exports, conversation archives, files, prompts, custom instructions, memory, saved memories, tools, GPTs, workflows, integrations, and relationships, then mapping as much as safely possible into Vellum primitives. Handles single-source and multi-source migrations with a unified, deduplicated inventory.

chatgpt-importSkill

Import conversation history from ChatGPT into Vellum

cli-discoverSkill

Discover which CLI tools are installed, their versions, and authentication status