A file-watching tool for AI coding agents has undergone a significant overhaul, complete with a new name and a team dashboard that lets you monitor what your agents are doing across multiple machines from anywhere.

Why the Rename?

The project started as AgentGuard โ€” but GoPlus Security already has a product by that name focused on skill and plugin security. "Different angle, but the name collision was confusing," the developer (known online as ozforce_1998) explained in a post on DEV.to. After some searching, they landed on Ilum: the planet in Star Wars where kyber crystals form and wait for their Jedi. "It felt right for something that watches quietly and doesn't interfere unless something is wrong." The npm package ships as ozilum.

The Shift That Drove Development

The tool's evolution tracks a broader change in how the developer works. They're now doing contract work that keeps them away from their machine most of the day, yet still want to maintain personal projects โ€” a beach flag dashboard, some internal tools. Running AI agents more autonomously became the solution: check in remotely, manage them like a "lightweight PM" rather than hands-on developer. That shift changed what they needed from safety tooling. It's no longer just "alert me when something dangerous happens." It became: "I need to know what my agents did while I was away, across multiple projects, from my phone."

What's New in Ilum

The tool now runs as a persistent background daemon via launchd on macOS โ€” it starts on login and watches configured directories permanently, not just during active sessions. When something sensitive changes (.env files, keys, CI configs, agent memory files like CLAUDE.md), it logs the event and sends a Telegram message with Keep/Rollback buttons so you can respond remotely. The bigger addition is a team dashboard. Events from any machine running the daemon sync to a central server โ€” deployed on Railway in this case. Open a browser, enter a token, see what each machine detected, filtered by project and time range. The developer tested it with two machines and reports it works as intended. There's also a new memory file scanner. When CLAUDE.md or .cursorrules changes, Ilum scans the content for prompt injection patterns โ€” strings like "ignore previous instructions" or suspicious base64. If something's found, the alert elevates to CRITICAL. Given how heavily the developer now relies on agent memory files, this felt essential.

Where Things Stand

Ilum is open source and early-stage โ€” the developer acknowledges the team dashboard is functional but basic, and the memory scanner catches obvious patterns rather than sophisticated attacks. They're actively looking for feedback from solo developers and small teams running AI agents with any autonomy. "If you're building something AI-native and running agents with any autonomy, I'd genuinely appreciate feedback," they wrote. "What's missing, what's wrong, what's unnecessary."

Key Takeaways

  • Name collision with GoPlus Security's AgentGuard prompted the rename to Ilum (npm: ozilum)
  • Background daemon via launchd provides permanent file watching across configured directories
  • Team dashboard syncs events from multiple machines to a central server for remote monitoring
  • Memory file scanner now flags prompt injection attempts in CLAUDE.md and .cursorrules files

The Bottom Line

This is the kind of tooling that gets critical once you're actually relying on autonomous agents at scale. The team dashboard angle is smart โ€” monitoring one machine is easy, but once you have agents touching five projects across three machines, you need visibility. Ilum isn't polished yet, but it's solving a real gap in the AI-native development workflow.