Security Lab — inside-the-loop.dev
Maintained by Aleksandr Krasnobai
DevSecOps × Secure AI Agents
https://security.inside-the-loop.dev
Engineering notes on secure AI runtime, agent control, and DevOps automation safety.
This is not a blog.
This is a working lab of minimal security artifacts.
Core Artifacts
Design Principles
- Default mode: read-only
- Tool usage is explicit and allowlisted
- Human approval for high-impact actions
- Secrets are isolated
- All actions are auditable
- Policy overrides content
Threat Model (Minimal)
Assume:
- External content is untrusted
- Alerts can be manipulated
- Docs can contain injection attempts
- Agents can be socially engineered
- Tool execution is the real risk surface
Secure AI Runtime.
https://security.inside-the-loop.dev