Detects and sanitizes prompt-injection and command-injection threats in AI agent inputs.
npx skills add https://github.com/zscole/AgentGuard --skill agent-guard
Please help me install this Agent Skill. Skill: agent-guard Source: https://github.com/zscole/AgentGuard/tree/main Command: npx skills add https://github.com/zscole/AgentGuard --skill agent-guard
Real-time detection and sanitization of prompt-injection and command-injection threats targeting AI agents, preventing unsafe prompts from altering behavior.
Install the skill in your OpenClaw or Claude MCP environment and start analyzing text with agent-guard analyze to detect threats.
High-intent search queries and answers about installing and using this skill.
You can detect prompt injection in AI agent inputs by running agent-guard analyze, which applies context-aware risk scoring to text and GitHub issues. It returns structured threat results, sanitized output with safe placeholders, and actionable recommendations.
Yes, you can sanitize command-injection threats in Claude MCP environments. Agent-guard integrates directly with Claude MCP servers to detect malicious inputs and replace threats with safe placeholders, preventing unsafe prompts from altering agent behavior.
No, you do not need external libraries for prompt-injection detection. Agent-guard's core logic requires only the Python standard library, though optional OpenClaw and Claude MCP integrations are available for seamless server deployment and threat analytics.
The best way to screen GitHub issues for AI security threats is using agent-guard to analyze issue content with context-aware risk scoring. It identifies prompt-injection and command-injection attacks, providing structured threat results and sanitizing unsafe inputs.
Yes, agent-guard works with OpenClaw for real-time AI agent security. It offers optional OpenClaw integration to detect and sanitize prompt-injection and command-injection threats, providing structured threat reporting and recommendations within the OpenClaw environment.
galeepParse FCS files and preprocess event data with Python.
seojoonkimDetect and block prompt-injection and secret-exfiltration attempts across multiple languages.
numbergroupDetect and sanitize dangerous prompts and commands for AI agents.
dexhunterScan installed skills and sanitize external input against injection attacks.
