Hi, I'm Ankang 👋

Open Source Developer · Computer Science Graduate Student · AI Agent Security Researcher

Building reliable software systems and exploring the security of AI agents.

I'm a Computer Science graduate student interested in software engineering, cybersecurity, distributed task systems, and AI agent security.

I enjoy turning ideas into practical, maintainable software — from backend architecture and plugin ecosystems to asynchronous task execution and security evaluation.

My current interests include:

• 🤖 AI Agent Security — runtime monitoring, behavior analysis, and security evaluation
• 🛡️ MCP & Tool Security — tool interfaces, permission boundaries, and secure execution
• ⚙️ Backend Engineering — plugin systems, task scheduling, workers, and asynchronous processing
• 🌍 Open Source — bug fixes, performance improvements, regression testing, and upstream contributions
• 🧠 AI & Machine Learning — LLM agents, deep learning, and evaluation methods

I value software that is not only functional, but also understandable, testable, and maintainable.

I contribute to open-source projects by identifying bugs, improving performance, strengthening validation, and making software systems more reliable.

perf(security): avoid Unicode folding for ASCII characters

• Optimized character handling to avoid unnecessary Unicode folding for ASCII characters.
• Improved the efficiency of a security-related character-processing path.
• Changed 2 files, adding 110 lines.
• Status: Merged into main on October 28, 2026.

View merged pull request →

The following PRs were verified as open on October 29, 2026. Their status may change as maintainers review and merge them.

I distinguish merged contributions from open pull requests so that every contribution is transparent and verifiable.

☁️ TFcloud Multi-platform Task Automation Platform A multi-platform, multi-account task automation system built around unified account management, scheduling, and plugin-based extensibility. Highlights 🔌 Plugin-oriented architecture ⏱️ Task scheduling and retries 🔄 Asynchronous task execution 👤 Multi-account management 🔐 Credential storage 🖥️ User and admin dashboards 🐳 Docker deployment 📊 Execution history and task status Stack PHP Webman React MySQL Docker

🛡️ agentSecure AI Agent Security Research & Learning A research-oriented repository exploring the engineering foundations of AI agent security monitoring and evaluation. Focus Areas 🤖 Agent behavior analysis 📝 Runtime behavior logging 🌐 Browser automation 🔧 Tool interface security 🐳 Isolated execution environments 🛡️ Agent security evaluation 🔎 Runtime security detection Stack Python FastAPI Playwright Docker

My research interests center on the security and reliability of autonomous AI agents.

AI Agent Security
│
├── Runtime Security Monitoring
│   ├── Behavior Analysis
│   └── Risk Detection
│
├── Tool & MCP Security
│   ├── Tool Invocation Validation
│   └── Permission Boundaries
│
├── Agent Interaction Security
│   ├── Multi-Agent Coordination
│   └── Authorization Consistency
│
└── Security Evaluation
    ├── Benchmark Design
    └── Reproducible Experiments

I'm particularly interested in this question:

How can we detect unsafe agent behavior early enough to prevent unauthorized actions and harmful outcomes?

I also explore how asynchronous execution, task provenance, and runtime controls affect the security and reliability of agent systems.

Readable        > Clever
Maintainable    > Temporary
Modular         > Coupled
Observable      > Invisible
Reliable        > "Works on my machine"

Good software should remain understandable and reliable as it grows.

I'm interested in collaborating on projects related to:

• AI Agents and Agent Security
• Cybersecurity and MCP Security
• Backend Architecture and Distributed Systems
• Automation Systems and Developer Tools
• Open Source Infrastructure

If you're working on something interesting, feel free to open an issue or start a discussion in one of my repositories.

Code · Research · Build · Repeat

_{Building things, breaking assumptions, and learning how systems really work.}