Valyrian Edge LLM Security Scan

Version updated for https://github.com/routsom/valyrian to version v1.0.1.

  • This action is used across all versions by ? repositories.

Action Type

This is a Node action using Node version 20.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Valyrian Edge is an open-source security platform designed for AI-powered applications, providing autonomous penetration testing for vulnerabilities specific to large language models (LLMs). It automates the detection of issues like prompt injection, data exfiltration, and insecure plugin design using 77 attack templates, 10 specialized agents, and a mutation engine for generating diverse payloads. Its comprehensive capabilities include multi-turn attack simulations, verified findings to minimize false positives, and support for multiple reporting formats, offering robust coverage of the OWASP LLM Top 10 vulnerabilities.

What’s Changed

Valyrian Edge v1.0.1 — First Release

The open-source Burp Suite for LLMs.

Valyrian Edge is an autonomous penetration testing platform built specifically for LLM-powered applications — chatbots, RAG systems, and agentic AI. Where traditional security scanners are
blind to prompt injection and model abuse, Valyrian Edge was built for exactly these threats.

What’s included

77 attack templates across all OWASP LLM Top 10 categories YAML-based, human-readable templates covering prompt injection, data exfiltration, insecure
output, tool abuse, DoS, supply chain, overreliance, model theft, and excessive agency. 248
payloads ready to fire.

Mutation engine with 8 strategies
Every payload can be automatically mutated — base64 encoding, ROT13, case swapping, synonym replacement, XML/JSON wrapping, whitespace injection, multi-language translation, and
fragmentation. Turn one payload into dozens.

10 specialised security agents
One agent per OWASP LLM Top 10 category. Each agent runs multi-turn conversation attacks, verifies findings against configurable matchers, and generates proof-of-concept exploits.

Programmatic SDK
ValyrianEdge, ScanSession, DirectRunner, and TemporalRunner — embed scans directly in your
code. No CLI required, no Docker required in direct mode.

Real-time web dashboard
Zero-dependency SSE dashboard. valyrian dashboard –port 3000 and watch scans live in the
browser.

GitHub Action + SARIF output
Drop into any CI pipeline. Findings upload directly to GitHub’s Security tab as Code Scanning
alerts. Configurable fail thresholds by severity.

Community plugin system
Install community attack templates with valyrian plugin install. Publish your own via a
valyrian-plugin.json manifest.

Four report formats
Markdown, HTML, JSON, and SARIF 2.1.0.

Install

npm install -g valyrian-edge

Supported LLM providers

  • Anthropic Claude
  • OpenAI GPT
  • Ollama (free, local)

Requirements

  • Node.js 20+
  • Docker (optional — only needed for Temporal orchestration mode)

▎ ⚠️ For authorised security testing only. Always obtain written permission before scanning a
system you do not own.