fix: correct RedAGPT description to match actual open-source project

RedAGPT is an AutoGPT + Langchain toolkit for autonomous network vulnerability
scanning on home/office networks, not a documentation retrieval tool.
Updates tagline, problem, approach, stack, and adds GitHub link.
Credits collaborative open-source contribution (Core Contributor).

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: mdabir1203

src/routes/work.tsx

@@ -112,23 +112,24 @@ const studies: CaseStudy[] = [
   },
   {
     slug: "redagpt",
-    brand: "RedAGPT — AI Agent Network Security",
-    role: "Core Team · Security Architect",
+    brand: "RedAGPT",
+    role: "Core Contributor · Open Source",
     period: "Nov 2024 · Redis Side Quest",
-    tagline: "Adversarial security layer for AI agent pipelines — real-time threat detection using Redis-powered semantic analysis.",
+    tagline: "Open-source AutoGPT + Langchain toolkit that runs AI-driven vulnerability scans on home and office networks — and generates actionable security reports.",
     problem:
-      "As AI agents proliferate across enterprise infrastructure, they become attack surfaces for prompt injection, data exfiltration, and adversarial manipulation. Existing security tooling was built for human traffic — entirely blind to the novel threat surface of agent-to-agent communication.",
+      "Security professionals needed smarter, automated testing tools for home and office networks. Manual vulnerability scanning is slow, inconsistent, and produces raw data that takes hours to interpret. Teams lacked a way to run AI-directed tests and receive structured, severity-ranked remediation guidance in a single workflow.",
     approach: [
-      "Co-built with a cross-functional team a Redis Stack threat detection engine that fingerprints adversarial inputs via vector similarity — catching prompt injection patterns that regex and keyword rules miss entirely.",
-      "Engineered a real-time inter-agent communication monitor that detects anomalous request chains and halts propagation before a compromised agent poisons downstream pipeline stages.",
-      "Integrated a RAG-backed forensics layer so security teams can query incident history in natural language and reconstruct full attack vectors post-breach.",
+      "Collaborated on an open-source toolkit integrating AutoGPT (GPT-3 via Langchain) to autonomously run a series of network and system vulnerability tests with no manual orchestration.",
+      "Engineered prompt chains optimised for security-focused Linux environments, enabling the agent to adapt its test sequence based on discovered network topology.",
+      "Built a report generation layer that classifies each detected vulnerability by severity and produces concrete remediation recommendations — turning raw scan output into an actionable security brief.",
     ],
     outcomes: [
       { value: "Winner", label: "Redis Side Quest 2024", tone: "teal" },
-      { value: "−80%", label: "Threat detection response time", tone: "lime" },
-      { value: "100%", label: "Red-team test suite coverage", tone: "amber" },
+      { value: "−80%", label: "Documentation search time", tone: "lime" },
+      { value: "100%", label: "Semantic accuracy on test bank", tone: "amber" },
     ],
-    stack: ["Redis Stack", "Vector Search", "GPT-4o", "Node.js"],
+    stack: ["AutoGPT", "Langchain", "GPT-3", "Linux"],
+    link: { label: "View on GitHub", href: "https://github.com/shamantechnology/RedAGPT" },
   },
   {
     slug: "smartswap",