OpenClaw
ClicheFactory skill for the OpenClaw agent framework — extraction tools with agent-level instructions.
What is OpenClaw
OpenClaw is an open-source framework for building AI agent workflows. The ClicheFactory skill gives an OpenClaw agent the ability to extract structured data from documents and convert them to markdown, with built-in instructions for handling errors and choosing the right extraction strategy.
Skill Installation
Install the ClicheFactory skill from the skill registry:
Or install from source:
The skill bundles agent instructions and the MCP server configuration. It is not a repeat of the MCP setup — it adds a layer of agent guidance on top.
Configuration
The skill needs the ClicheFactory MCP server registered with OpenClaw. If you haven't already:
"command": "uvx",
"args": ["clichefactory-mcp"],
"env": {
"LLM_MODEL_NAME": "gemini/gemini-3-flash-preview",
"LLM_API_KEY": "your-key"
}
}'
See MCP Server → OpenClaw Setup for details on local vs. service mode configuration.
Usage
Once installed, the agent follows a structured workflow when asked to extract data from documents:
- Inspect the document — the agent calls
to_markdownto understand the document structure and content. - Build a schema — based on the document content and the user's request, the agent constructs a JSON schema describing the fields to extract.
- Extract — the agent calls
extractwith the document and schema, choosing the appropriate extraction mode. - Handle errors — if extraction fails or returns validation errors, the agent adjusts the schema or mode and retries.
- Diagnose — if repeated failures occur, the agent calls
doctorto check the configuration and reports findings to the user.
The skill instructions guide the agent through this workflow automatically — you just ask for what you need in natural language.
Supported File Types
The ClicheFactory skill supports PDF, images (PNG, JPG, WebP, GIF, BMP), DOCX, XLSX, CSV, EML, TXT, and more — see Supported File Types for the full list.