Openclaw Skills' Deep Doc Processor turns PDFs, Word files, Markdown, web pages, and text into layered summaries, comparison reports, and deep Q&A.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install deep-doc-processor
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install deep-doc-processor using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Deep Doc Processor?
Openclaw Skills Deep Doc Processor is a document intelligence skill built for long-form reading, information extraction, and synthesis across PDF, Word, Markdown, TXT, web pages, and other text sources. It automatically adapts when a user asks to summarize, extract key points, compare multiple sources, or probe a document with follow-up questions.
Designed for Openclaw Skills workflows, it selects the right reader for each source, then produces layered summaries, structured analysis reports, and precise evidence-based answers. It also preserves source files unchanged, writes outputs to the output directory, and redacts sensitive personal information in generated reports.
Deep Doc Processor Use Cases
- Use Openclaw Skills to summarize a lengthy PDF, Word document, Markdown file, or plain-text note.
- Extract core arguments, evidence chains, contradictions, and decision risks from dense documents.
- Compare multiple source documents side by side and surface agreement, divergence, and data gaps.
- Generate a structured report with core insights, key data, action recommendations, and risk notes.
- Answer follow-up questions about a document with source-localized interpretation and direct evidence.
- Analyze web pages or OCR text from images when the content is not available as a local file.
How Deep Doc Processor Works
- Detect a trigger such as summarize, extract key points, compare, analyze, or deep question, then load the skill automatically.
- Choose the correct ingestion path:
file-agentwithread_filefor PDF, DOCX, XLSX, and PPTX;read_textfor TXT, Markdown, or code;web_fetchorbrowser-agentfor URLs; andanalyze_imagefor OCR. - Normalize the content and segment it into reading units for layered analysis, including one-sentence summary, paragraph summary, outline, claims, evidence, causal links, contradictions, and risks.
- If multiple sources are provided, build a cross-document comparison matrix to identify shared conclusions, key differences, and evidence gaps.
- If the user requests a report or follow-up Q&A, generate a structured Markdown deliverable with cited evidence, action items, and source-localized explanations.
- Write the final artifact to the output directory, keep originals unchanged, and redact sensitive personal data in the generated output as
***.
Deep Doc Processor Setup
Openclaw Skills Deep Doc Processor does not require a standalone package install; it is activated by the agent when the trigger matches. To make sure it works reliably, place the skill in your skills directory, create an output folder, and confirm the required tools are available.
mkdir -p ~/.openclaw/skills/deep-doc-processor
cp SKILL.md ~/.openclaw/skills/deep-doc-processor/
mkdir -p output
Then verify the agent can access file-agent, web_fetch, browser-agent, analyze_image, and write permissions for output/. If you are running in a managed Openclaw Skills environment, confirm document-reading and OCR tools are enabled before testing with a summary or comparison prompt.
Deep Doc Processor Data Schema & Taxonomy
Input sources
| Source type | Reader | Normalized output |
|---|---|---|
| PDF / DOCX / XLSX / PPTX | file-agent -> read_file |
Extracted text plus document structure |
| TXT / Markdown / code | read_text or file-agent |
Raw text with line structure preserved |
| Web page URL | web_fetch or browser-agent fallback |
Page content converted to readable text |
| Image with text | analyze_image |
OCR text plus inferred layout cues |
Analysis taxonomy
summary: one-sentence summary, paragraph summary, and full outlineinsights: key viewpoints, argument chains, evidence, and contradictionscomparison: dimension matrix, consensus findings, and divergence pointsreport: action recommendations, risk notes, and key data tablesqa: source-localized explanations with quoted evidence when needed
Output artifacts
- Markdown analysis reports written to the output directory
- Multi-document comparison reports when two or more sources are supplied
- Redacted sensitive values replaced with
*** - Metadata footer containing generation time and processed source path or URL
- Original files remain unchanged; only derived artifacts are written
Deep Doc Processor Advanced Features
- Automatic trigger loading for summarize, extract, compare, report, and deep-question prompts.
- Multi-format ingestion across PDFs, Office files, Markdown, TXT, web pages, and OCR images.
- Layered summarization from one-line gist to paragraph-level synthesis and full outlines.
- Evidence-aware extraction of claims, arguments, contradictions, causal relationships, and risk signals.
- Multi-document cross comparison with difference matrices and consensus extraction.
- Structured Markdown report generation for executive summaries and decision support.
- Sensitive-data redaction (
***) and non-destructive output handling. - Precision follow-up Q&A with source-localized answers and direct quotations.
- Built to fit Openclaw Skills workflows for repeatable document intelligence automation.
SKILL.md
Loading
Related Openclaw Skills
Wool Hunter (薅羊毛助手)
A unified shopping and deal aggregator designed to help users programmatically discover coupons, compare prices, and query welfares across multiple platforms.
PDF Label - Made In China
Batch-inserts "Made In China" into Amazon FBA label PDFs with barcode-aware positioning, test-page previews, and mixed-SKU support.
Pixel Asset Compiler
A utility to compile AI-generated pixel character reference sheets and individual actions into validated, game-ready assets and engine scenes.
Save Claude Notes
Save Claude Notes captures Claude Code Q&A into Markdown notes with transcript-aware defaults and flexible save targets.
Douyin Video Watermark Removal & Summary
Converts a Douyin link into a watermark-free video, extracted audio, transcript, and a structured summary in one Openclaw Skills pipeline.
Charlotte On The Cheap MCP
A read-only Charlotte, NC discovery skill that finds free and cheap events, deals, and local guides with date-aware search and full article lookup.
Featured*
