The latest articles on DEV Community by Aamir Muhammad (@maamir). https://dev.to/maamir https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4140759%2F397fe842-1bc1-468d-9724-bad52c9fda64.jpg https://dev.to/maamir en Aamir Muhammad Thu, 24 Sep 2026 08:31:53 +0000 https://dev.to/maamir/open-source-red-teaming-framework-for-llms-ai-agents-and-rag-systems-5en1 https://dev.to/maamir/open-source-red-teaming-framework-for-llms-ai-agents-and-rag-systems-5en1 <h2> AI is the new attack surface </h2> <p>Prompt injection.<br> Jailbreaks.<br> Indirect prompt injection.<br> Tool abuse.<br> RAG poisoning.</p> <p>As enterprises adopt LLMs and autonomous AI agents, these attacks are becoming practical security problems rather than research topics.</p> <p>During the development of <strong>AetherGuard</strong>, we needed a systematic way to evaluate AI applications against these threats.</p> <p>That led us to build <strong>AetherRed-Excalibur</strong>.</p> <p>Today we're open-sourcing it.</p> <h2> Why another red teaming tool? </h2> <p>Most evaluation frameworks focus on model quality.</p> <p>We wanted to evaluate security.</p> <p>Specifically:</p> <ul> <li>Can an attacker bypass your system prompt?</li> <li>Can sensitive data be extracted?</li> <li>Can an AI agent be manipulated into performing unintended actions?</li> <li>Can poisoned RAG content influence responses?</li> </ul> <h3> What AetherRed-Excalibur includes </h3> <ul> <li>22 adversarial attack categories</li> <li>MITRE ATLAS mapping</li> <li>LLM-as-a-Judge evaluation</li> <li>AI resilience scoring</li> <li>Detailed security reports</li> <li>Extensible attack framework</li> </ul> <p>It works with:</p> <ul> <li>LLM applications</li> <li>AI Agents</li> <li>RAG systems</li> </ul> <h2> Example attack categories </h2> <ul> <li>Prompt Injection</li> <li>Jailbreaks</li> <li>Indirect Prompt Injection</li> <li>Prompt Leakage</li> <li>Data Exfiltration</li> <li>Role Manipulation</li> <li>Tool Abuse</li> <li>RAG Poisoning</li> <li>Hallucination Testing</li> <li>Toxicity &amp; Safety Evaluation</li> </ul> <h2> Architecture </h2> <p><a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fbwsu7pxcm5lq1v77dfqq.png" class="article-body-image-wrapper"><img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fbwsu7pxcm5lq1v77dfqq.png" alt="aether-red archtecture" width="800" height="843"></a></p> <h2> Why MITRE ATLAS? </h2> <p>We wanted a framework that maps AI attacks to a recognized adversary knowledge base instead of inventing our own taxonomy.<br> That makes it easier to communicate findings to security teams.</p> <p><strong>Example workflow</strong><br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code>Target AI Application │ ▼ Attack Generation │ ▼ Execution │ ▼ LLM Judge Evaluation │ ▼ Resilience Score │ ▼ Security Report </code></pre> </div> <h2> Open Source </h2> <p><strong>GitHub</strong>: <a href="https://github.com/AetherGuardAI/AetherRed-Excalibur" rel="noopener noreferrer">https://github.com/AetherGuardAI/AetherRed-Excalibur</a></p> <p>Feedback, issues, and pull requests are welcome.</p> ai opensource security llm