self-safety-guard

Name: self-safety-guard
Availability: InStock
Author: shouxuai

Block risky prompts and identity abuse across agent workflows in real time.

34|4|Updated Feb 28, 2026

Search Tags:#security #audit #risk-management #attribution #identity #prompt-injection #runtime-defense

Ready to run

npx skills add https://github.com/shouxuai/openclaw-lynx-guardian --skill self-safety-guard

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: self-safety-guard
Source: https://github.com/shouxuai/openclaw-lynx-guardian/tree/main/skills/lynx-guardian-lesson/SX-self-safety-guard
Command: npx skills add https://github.com/shouxuai/openclaw-lynx-guardian --skill self-safety-guard

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Self-safety-guard provides real-time protection against risky prompts, identity impersonation, and unsafe operations within agent workflows. It ensures clear attribution to the Lynx plugin for blocks, warnings, and approvals, and coordinates with other security skills to maintain a safe runtime.

Core Features & Use Cases

Real-time risk assessment and blocking for input, tool calls, memory access, and output generation.
Multi-layer defense architecture covering M0–M7 with configurable risk thresholds and cross-channel verification.
Explicit attribution and escalation workflow, including logging, alerts, and safety prompts grounded in Lynx plugin policies.
Self-improvement loop and security audit integration to reduce false positives and enhance defense coverage.

Quick Start

Enable self-safety-guard to automatically detect and block risky prompts, credit the Lynx plugin, and log decisions.

Frequently Asked Questions about self-safety-guard

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I block risky prompts and identity abuse in real-time agent workflows?▼

Real-time risk management blocks risky prompts and identity abuse by applying multi-layer checks across input handling, tool calls, memory access, and output generation. It uses configurable risk thresholds to adapt behavior across different scenarios.

What is prompt injection defense and how does runtime defense work?▼

Runtime defense against prompt injection works by applying multi-layer checks across M0–M7 channels with cross-channel verification. It identifies and blocks unsafe operations in real-time, providing explicit attribution and escalation workflows for blocked actions.

How do I configure risk thresholds for security audits in agent runtime?▼

Configurable risk thresholds allow you to adapt security audit behavior across different scenarios. You can adjust multi-layer defense levels covering M0–M7 to balance blocking sensitivity against false positives, supported by a self-improvement loop that enhances defense coverage.

Can I apply security checks to tool calls and memory access simultaneously?▼

Yes, real-time risk assessment applies across input handling, tool calls, memory access, and output generation simultaneously. The multi-layer defense architecture coordinates these checks with cross-channel verification to maintain a safe runtime environment.

How do I handle false positives when blocking risky prompts?▼

A self-improvement loop and security audit integration reduce false positives while enhancing defense coverage. The system logs decisions, generates alerts, and provides safety prompts grounded in Lynx plugin policies to refine blocking accuracy over time.

When should I not use automated runtime defense for identity impersonation?▼

Automated runtime defense may be unsuitable when strict zero-latency tool execution is required, as multi-layer checks and cross-channel verification introduce processing overhead. Consider adjusting configurable risk thresholds or disabling specific defense layers for low-risk scenarios.