Detect and block prompt-injection and secret-exfiltration attempts across multiple languages.
npx skills add https://github.com/seojoonkim/prompt-guard --skill prompt-guard
Please help me install this Agent Skill. Skill: prompt-guard Source: https://github.com/seojoonkim/prompt-guard/tree/main Command: npx skills add https://github.com/seojoonkim/prompt-guard --skill prompt-guard
Prompt Guard protects AI agents from prompt injection and secret-exfiltration attempts by providing multi-language detection, decoding pipelines, and enterprise-grade DLP.
To get started, run a prompt-guard scan on a user message to determine if it should be allowed or blocked.
High-intent search queries and answers about installing and using this skill.
Prompt-injection defense works by applying a decode-then-scan pipeline and language detection to user messages. It classifies threats and outputs SHIELD.md compliant decisions to block dangerous prompts or allow safe ones in production.
Yes, you can prevent secret-exfiltration by deploying this to redact credentials in LLM outputs. It applies enterprise-grade DLP, uses canary tokens for tracking unauthorized access, and blocks exfiltration attempts across multiple languages.
Yes, prompt-injection detection supports multiple languages including EN, KO, JA, ZH, RU, ES, DE, FR, PT, and VI. This multi-language coverage ensures pervasive attacks are caught regardless of the input language.
You can integrate threat intelligence for rapid pattern updates using the HiveFence network integration. This allows operators to share threat intelligence and receive rapid updates to keep AI agents protected against new attacks.
To set up prompt-injection defense and JSONL logging, you need the pyyaml dependency installed in your environment. The system then supports rate limiting, canary tokens, and JSONL logging to track and manage detected threats.
The best way to automate protection for a group assistant is deploying a SHIELD.md compliant defense pipeline. It automatically classifies threat levels, applies rate limiting, and outputs automated block or allow decisions for production environments.
lpjhelderInstall and configure kronon-forge with Node.js and npm.
chenchunrunExtract MITRE ATT&CK techniques from security reports and generate Sigma rules.
Happy-Technologies-LLCCorrelate security incidents with events, vulnerabilities, and threat intelligence.
ramirlmDetect prompt injection and block secret exfiltration across multiple languages.
