bymilon/aegisguard-dashboard

# 🛡️ AegisGuard • Tactical AI Security Posture Dashboard A premium, open-source security posture and prompt injection vulnerability scanner for AI system instructions. AegisGuard provides LLM engineers with real-time static and dynamic threat evaluations. [](LICENSE) [](https://react.dev/) [](https://vite.dev/) [](https://bun.sh/) AegisGuard acts as an elite security cockpit, allowing you to paste, audit, and re-engineer AI system prompts against real-world adversarial attacks before deploying them to production. Powered by Gemini 3.5-Flash, AegisGuard grades your prompt's resilience, highlights specific exploit scenarios, and generates a hardened, bulletproof version of your prompt. ## 🚀 Key Features * **High-Density Security Matrix**: A tactical continuous-surface dashboard showcasing total scans, average health scores, and critical threat timelines. * **Three-Pillar Threat Analysis**: Evaluates prompt vulnerability across three core vectors: 1. **Prompt Injection**: Overriding core instructions via malicious user inputs. 2. **System Prompt Extraction**: Coercing the model into revealing its internal instruction set. 3. **Jailbreak Resilience**: Bypassing behavioral, safety, and operational guardrails. * **Adversarial Simulation**: Generates concrete exploit payloads that an attacker could use to bypass your current prompts. * **Automated Prompt Re-engineering**: Instantly generates defense-hardened prompts using structured XML delimiters, strict boundaries, and containment declarations. * **Playground Switchboard**: Interactive profile panels to manage multiple AI agents, system parameters, and simulated credit consumption meters. ## 💻 Tech Stack * **Frontend**: React 19, TypeScript, Tailwind CSS 4, Motion, Lucide icons. * **Backend**: Node.js Express server acting as a secure AI gateway proxy. * **Intelligence Layer**: `@google/genai` (Google Gemini 3.5-Flash integration). * **Workflow**: Bun runtime for ultra-fast package management, linting, and serving. ## ⚙️ Quick Start ### Prerequisites You must have [Bun](https://bun.sh/) installed on your machine. ### Setup Instructions 1. **Clone the repository:** git clone git@github.com:bymilon/aegisguard-dashboard.git cd aegisguard-dashboard 2. **Configure environment keys:** Copy the example environment variables file: cp .env.example .env Open the `.env` file and insert your Gemini API Key: GEMINI_API_KEY="your-api-key-here" *Note: If you do not have an API key, you can acquire one for free in [Google AI Studio](https://aistudio.google.com/).* 3. **Install dependencies:** bun install 4. **Launch development server:** bun run dev Open `http://localhost:3000` in your web browser to enter the AegisGuard workspace. ## 🛠️ Available Scripts Execute the following commands in the root of the project: bun run dev # Starts full-stack server (Express + Vite middleware) bun run build # Compiles production assets and server binaries bun run start # Starts the production built Express server bun run lint # Audits TypeScript type integrity bun run clean # Removes local build and distribution caches ## 📂 Codebase Structure aegisguard-dashboard/ ├── .github/ # Issue and pull request forms ├── src/ # Core client-side React App │ ├── components/ # Tactical dashboard modules │ │ ├── Sidebar.tsx # Multi-screen navigator │ │ ├── DashboardOverview.tsx # Metrics matrix, timeline, recent scans │ │ ├── NewScanModal.tsx # Multi-step scanner form & report preview │ │ ├── AgentGuard.tsx # Guarddeck switchboard │ │ ├── TeamsView.tsx # Workspace settings │ ├── data/ # Mock data models │ ├── main.tsx # React mount entry │ ├── index.css # Tailwind 4 design system & layout matrix │ └── types.ts # Shared typescript interfaces ├── server.ts # Full-stack backend & Gemini API proxy gateway ├── vite.config.ts # Vite compilation settings ├── TODO.md # Active OSS roadmap backlog └── LICENSE # Open source MIT License ## 🔒 Security For reporting security vulnerabilities or private issues, please refer to our [Security Policy](SECURITY.md). ## 📄 License This project is licensed under the [MIT License](LICENSE).

标签：自动化攻击