Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
Initial public release of Tork Guardian 1.0.0. - Adds comprehensive governance and network security layer for OpenClaw agents. - Introduces policy enforcement (PII redaction, shell/file/network controls) with configurable tiers (minimal, development, production, enterprise). - Integrates standalone security scanner CLI and programmatic API with risk scoring, compliance receipts, and Tork Verified badges. - Includes pre-built configs, standalone utility functions, and full TypeScript/JS support. - Extensive handling and logging for LLM requests, tool calls, and network activity. - Removed obsolete and relocated scanner files and test configs for streamlined structure.
What This Skill Does
Security and governance layer for OpenClaw agents. Handles PII redaction, tool access control, network policy enforcement, and compliance receipt generation. Includes a pre-installation vulnerability scanner that checks skills against 14 security patterns.
Combines PII redaction, tool governance, network controls, and pre-install vulnerability scanning in a single package rather than requiring separate solutions for each concern.
When to use it
- Redacting emails and phone numbers from user messages before they reach an LLM
- Blocking destructive shell commands like rm -rf from agent tool execution
- Scanning a third-party skill for vulnerabilities before installing it
- Enforcing strict network policies to prevent SSRF attacks in production deployments
- Generating compliance receipts for audit logs when agents process sensitive data
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Scan the ./payment-processor skill for security issues before I install it
- 1Initialize SkillScanner from @torknetwork/guardian
- 2Run scanner.scanSkill('./payment-processor') to analyze the skill directory
- 3Read the risk score (0-100) and verdict from the report
- 4Generate a Tork badge based on the verdict
- 5Return findings with severity levels and recommended action
Risk score report showing verdict ('verified', 'reviewed', or 'flagged'), individual findings by category, and a badge markdown string for the README
Requirements
Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.