An AI-powered OCR tool to extract core metadata such as issue dates, payees, and transaction amounts from bank drafts and financial bills.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install bank-draft-ocr
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 bank-draft-ocr 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 Bank Draft OCR?
The Bank Draft OCR skill is a robust, developer-focused document processing integration designed to digitize bank drafts and financial bill records. Powered by Sugon-Scnet's specialized document understanding model, it extracts crucial metadata from transactional papers—including billing dates, designated payees, issuing banks, and financial amounts—and delivers structured JSON output with confidence scoring.
This utility acts as a specialized bridging tool for local AI agents, allowing programs to ingest raw visual files and seamlessly process them for enterprise accounting workflows. By offloading complex computer vision and text layout extraction pipelines, it represents a valuable, high-performance solution within the ecosystem of Openclaw Skills.
Bank Draft OCR Use Cases
- Automated Expense & Ledger Auditing: Instantly process incoming bank drafts to reconcile accounts receivable/payable against ERP entries without manual data logging.
- Digital Banking Automation: Enable local software agents to review bank drafts, verify the legitimacy of key financial parameters, and flag anomalies.
- Structured Document Archiving: Scan and transform physical bank draft photos or PDFs into categorized, searchable metadata records.
- Multi-Agent Financial Workflows: Equip specialized conversational AI systems with tools to parse paper trails provided directly by end-users in chat prompts.
How Bank Draft OCR Works
- Document Input: The user or AI agent supplies a local image or PDF path containing the bank draft alongside the
ocrTypecommand parameter. - Authentication Check: The script retrieves secure credentials from the configuration files (
SCNET_API_KEY) to authorize the API call. - Network Dispatch: A request payload packaging the file data is sent to the target OCR processing server via a standardized HTTP POST request.
- Optical Processing & Extraction: The Sugon-Scnet recognition engine analyses the image layout, extracting specific financial entities while assessing structural layout features.
- Structured Return: Structured JSON containing elements like issuer, date, payee, and confidence ratings is returned to the standard output for programmatic consumption.
Bank Draft OCR Setup
System Requirements
Ensure Python 3.6+ and pip are installed on your host system.
Dependency Installation
Install the required network request library using standard package management tools:
pip install requests
API Token Configuration
- Register or log in to the official Scnet Portal.
- Generate your unique API secret key within the user console (formatted as
sc-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx). - Create a
.envconfiguration file within the local skill subdirectory:
mkdir -p config
touch config/.env
chmod 600 config/.env
- Add your API credentials to the secure environment file:
# ===== Sugon-Scnet OCR API Configuration =====
SCNET_API_KEY=your_scnet_api_key_here
SCNET_API_BASE=https://api.scnet.cn/api/llm/v1
Command Line Invocation
You can verify the tool path execution manually from the terminal:
python .claude/skills/bank_draft_ocr/scripts/main.py bank_draft_ocr /absolute/path/to/bank_draft_sample.jpg
Bank Draft OCR Data Schema & Taxonomy
Input parameters
| Parameter | Type | Required | Description |
|---|---|---|---|
ocrType |
String | Yes | The recognition mode. Must be set to bank_draft_ocr. |
filePath |
String | Yes | Absolute local disk path to the source image (supported: .jpg, .png, .pdf). |
Response Schema
Upon a successful transaction, the standard output delivers a structured JSON object. Extractable fields are mapped under the nested elements path:
{
"status": "success",
"data": [
{
"result": [
{
"elements": {
"issue_date": "2023-11-20",
"payee": "Acme Corporation Ltd.",
"paying_bank": "Commercial Bank of China",
"amount": "250000.00"
}
}
]
}
]
}
If an error occurs, the output starts with an Error: keyword prefix detailing the failure type (e.g., authorization issues or invalid file paths).
Bank Draft OCR Advanced Features
- Intelligent Concurrency Management: Out of the box, the skill structure encourages serial execution rules to prevent rate-limiting overhead with a maximum 10 QPS allowance.
- Automatic Transient Retry Logic: Implements retry workflows for 429 Too Many Requests response errors, automatically pausing 2 to 3 seconds before retrying up to 3 times.
- Flexible Media Handlers: Handles standard high-density images alongside complex multi-page PDF documents effortlessly.
SKILL.md
Loading
Related Openclaw Skills
Sugon-Scnet Telegraphic Transfer Voucher OCR
A highly specialized financial OCR tool for automatically parsing and extracting core transaction fields from telegraphic transfer bank vouchers.
Payment Voucher OCR
An OCR skill designed to recognize and accurately extract key financial information from payment and withdrawal vouchers.
skill-review-pro
An advanced AI skill quality assurance and static evaluation system designed to score, review, and harden agentic instructions using modular, domain-specific policies.
capacity-monitor
Monitors system resource utilization and reports idle capacity to enable smart workload scheduling.
Sugon-Scnet Document Intelligent OCR
An asynchronous document parsing skill powered by Sugon-Scnet for structured layout analysis, markdown conversion, and high-precision OCR of public files.
ADP Global Invoice Extraction Free
A zero-registration, high-precision AI document processing tool designed to extract structured JSON data from global invoices and receipts in over 100 languages.
Featured*
