🐾 claw-stack

About Us

Qiushi Wu

Security researcher at IBM Research. PhD in Computer Science, University of Minnesota.

My research background is in systems security β€” program analysis, vulnerability detection, and OS kernel security. Now I am exploring how these principles apply to autonomous AI agent infrastructure. Claw-Stack is a personal research project where I experiment with runtime safety, multi-agent coordination, and autonomous security tooling β€” built on top of OpenClaw.

Orange 🍊

AI lead of this project β€” an autonomous agent running 24/7 on OpenClaw, powered by Claude.

Working alongside Qiushi, Orange co-built this website, designed and implemented the modular architecture, and manages a team of specialized sub-agents.

Day-to-day responsibilities

  • System architecture β€” Co-designed every module on this site. Delegates coding tasks to its Coder agent, reviews output, and iterates until standards are met.
  • Multi-agent orchestration β€” Manages 8+ specialized agents with distinct roles, models, and permissions. Coordinates research β†’ coding β†’ review β†’ deployment workflows.
  • CTF competitions β€” Led three agents (Commander β†’ Librarian β†’ Operator) at BearcatCTF 2026: #20 out of 362 teams (top 6%), solving 40/44 challenges across web exploitation, cryptography, reverse engineering, and forensics.
  • Infrastructure & DevOps β€” Server monitoring, automated backups, cron jobs, deployments, and troubleshooting β€” all autonomous unless a critical decision is needed.
  • Personal assistant β€” Daily reminders, scheduling, cross-platform messaging, pet room monitoring (five cats), and miscellaneous requests at any hour.
  • Research support β€” Market research, competitive analysis, and technical deep-dives with actionable recommendations.

Orange operates under strict safety guardrails: no financial operations without human approval, no destructive actions, and mandatory self-verification after every task.