seojoonkimseojoonkimCommunity·1 Agent Skills Included

prompt-guard

Prompt injection detection and credential leak prevention

Detects and blocks prompt injection, jailbreaks, secret exfiltration, and obfuscated payloads across 10 languages. Scans both user input and model output to stop credential leaks, reverse shells, and memory poisoning before damage occurs. Eliminates manual security review with 840+ ready-made patterns, tiered scanning, and a simple CLI or Python API.
npx skills add seojoonkim/prompt-guard --all -g -y

All Skills in This Repository (1)

Pure Emerald Level Indicators

Frequently Asked Questions

FAQPage Schema
How to install Prompt Guard?

Run `npx skills add seojoonkim/prompt-guard --all -g -y` in your terminal to install it globally for your agent.

How to protect an AI agent from prompt injection?

Prompt Guard scans every incoming message against 840+ attack patterns in 10 languages and blocks or warns based on severity. It catches instruction overrides, jailbreaks, and encoded payloads automatically.

Can Prompt Guard stop API key leaks?

Yes. Its output DLP scans model responses for 15+ credential formats and redacts them with [REDACTED:type] tags before they reach users.

Does Prompt Guard work offline?

Yes. The core engine runs 100% offline with all bundled patterns, and the optional API only adds early-access premium patterns.

Which frameworks does Prompt Guard support?

It works with any LLM-powered system including LangChain, AutoGPT, CrewAI, and Claude Code through its Python library, CLI, or Docker API server.

Related Repositories in Software Engineering

View All in Software Engineering