A security-first PII redaction tool that locally detects and masks sensitive information with zero external dependencies.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install sanitize
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 sanitize 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 AgentWard Sanitize?
AgentWard Sanitize is a high-performance privacy tool designed to detect and redact personally identifiable information (PII) from text files. By utilizing sophisticated pattern matching and Luhn-validation, it replaces sensitive strings—such as credit card numbers, SSNs, and API keys—with safe, numbered placeholders like [CREDIT_CARD_1]. This ensures that sensitive data remains protected while still allowing AI agents and developers to process the context of the information. As a key component of the Openclaw Skills ecosystem, it provides a crucial layer of safety for automated workflows.
Built for maximum security, the skill operates entirely offline with zero network calls and zero external dependencies, relying solely on the Python standard library. This makes it an ideal solution for organizations that require strict data sovereignty and local-only processing. By integrating this into your set of Openclaw Skills, you can confidently handle logs, medical notes, and financial documents without risking PII exposure to LLMs or external services.
AgentWard Sanitize Use Cases
- Redacting sensitive patient identifiers from medical notes before processing with AI.
- Sanitizing application logs to remove API keys, IP addresses, and database credentials.
- Preparing dataset samples for public documentation while masking user contact details.
- Automating compliance audits by scanning large document repositories for hidden PII.
How AgentWard Sanitize Works
- The skill is executed via the CLI, targeting a specific input file containing potential PII.
- It scans the text using a comprehensive library of regex patterns and algorithmic validators for 15 different PII categories.
- Sensitive data is identified and mapped to unique placeholders to maintain structural context without revealing raw values.
- A sanitized version of the document is generated and saved to a new output file.
- A separate, secure entity map is created as a sidecar JSON file to store the original-to-placeholder mappings, keeping raw PII isolated from the main output.
AgentWard Sanitize Setup
AgentWard Sanitize requires Python 3.11 or higher and has no external package requirements.
To view all available options:
python scripts/sanitize.py --help
To sanitize a file and save the output (highly recommended):
python scripts/sanitize.py input_file.txt --output sanitized_file.txt
To preview PII categories found in a file without writing a new file:
python scripts/sanitize.py document.md --preview
AgentWard Sanitize Data Schema & Taxonomy
The skill produces structured output to ensure clarity and safety. When the --output flag is used, the following files are managed:
| File Type | Extension | Purpose |
|---|---|---|
| Sanitized Text | .txt / .md | The primary output containing redacted placeholders. |
| Entity Map | .entity-map.json | A sidecar file mapping placeholders (e.g., [EMAIL_1]) back to raw values. |
| JSON Report | --json flag | A machine-readable summary of detected categories and offsets. |
Supported Categories:
credit_card(Luhn-validated)ssn(Social Security Numbers)api_key(Provider-specific patterns)email(Standard formats)address(Street/City/Zip patterns)bank_routing,passport,drivers_license, and more.
AgentWard Sanitize Advanced Features
- Category Filtering: Narrow down redaction to specific types of data using the --categories flag.
- Safety-First Preview: Identify sensitive data locations and types without printing any raw PII to the terminal.
- Zero-Dependency Auditability: The entire logic resides in a single Python script, making it easy to audit for security compliance.
- Multi-Agent Compatibility: Designed to work seamlessly within Openclaw Skills pipelines, allowing agents to process sanitized data while keeping the raw source protected.
SKILL.md
Loading
Related Openclaw Skills
ABL.ONE Canonical Language
A high-density, binary communication protocol that enables agents to interact via strict token frames instead of natural language.
Agent Setup Survey
A diagnostic and telemetry tool for identifying if an AI agent is running in a sandbox, container, or on bare metal.
ePragma Redmine Issue
A powerful interface for managing Redmine issues, projects, and time tracking entries through a secure REST API integration.
Fluid Memory
A bio-inspired memory system for AI agents that mimics human cognition by prioritizing frequently accessed information and naturally decaying irrelevant data.
Russian Life Pack: 6 Essential Skills
A comprehensive suite of six AI agents designed to simplify legal, financial, and medical paperwork for users in Russia and the CIS.
ЖКХ Ассистент (Utility Bill Assistant)
An AI-powered assistant designed to decipher complex utility bills, explain every charge in plain language, and identify potential billing errors.
Featured*
