Imran Siddique

Chief platform officer @ OPAQUE Systems. Tweets personal.

[Imran Siddique @mosiddi]

OPAQUE Extends the Agent Governance Toolkit with Verifiable Identity and First Ever Verifiably Governed and Secure MCP https://t.co/RLnPOzTyh7 via @Street_Insider

[Imran Siddique @mosiddi]

Check out my latest article: Policy Is Not Proof https://t.co/w7NHQT6Jqw via @LinkedIn

[Imran Siddique @mosiddi]

Check out my latest article: From Software to Silicon: What AGT Taught Me About the Real Ceiling in AI Governance https://t.co/finxYHi6Ur via @LinkedIn

[Microsoft Learn @MicrosoftLearn]

Runtime governance is quickly becoming part of the AI agent stack.
This blog breaks down how Agent Governance Toolkit (AGT) + Microsoft Agent Framework (MAF) help developers build governed agents with built-in policy enforcement, sandboxing, auditability, and cost controls. Read here: https://t.co/cGi24V50J4

[Imran Siddique @mosiddi]

Governing AI Agents Against Every OWASP Agentic Risk: A Deep Dive with the Agent Governance Toolkit | Microsoft Community Hub https://t.co/7OKYip6twC

[Imran Siddique @mosiddi]

Ubuntu Summit 26.04 | Day Two | AGT https://t.co/rauAwX8On0 via @YouTube

#AGT #AgentGovernanceToolkit #AgentOS #AgentMesh

[Imran Siddique @mosiddi]

AGT + Atbash AI = Awesome governance!

[Yosef Soso @0x50so]

Was an honor spending time with you!

[Imran Siddique @mosiddi]

Applying Site Reliability Engineering to Autonomous AI Agents | Microsoft Community Hub https://t.co/h7KHCn1Zvt

[Imran Siddique @mosiddi]

Governance at the Speed of Agents: Microsoft Agent Framework and Agent Governance Toolkit, Better Together https://t.co/v3aKuXRJin

[Imran Siddique @mosiddi]

Check out my latest article: 20 Hard Questions About AI Agent Governance That Nobody Is Asking https://t.co/VxhcxB5KTG via @LinkedIn

[Imran Siddique @mosiddi]

Check out my latest article: 20 Hard Questions About AI Agent Governance That Nobody Is Asking https://t.co/HyER6uzUHb via @LinkedIn

[Imran Siddique @mosiddi]

We just released a deep dive into Shift-Left Governance for AI agents over at the Microsoft Community Hub. The core idea is simple: the earlier you catch a violation, the cheaper it is to fix. The Agent Governance Toolkit (AGT) allows you to bake compliance—like dependency pinning and policy validation—directly into the developer workflow. If you are moving away from code-centric development toward orchestration-centric systems, this is the infrastructure you need to keep things from breaking at scale. https://t.co/BTHgSk5R8L

[Imran Siddique @mosiddi]

Introducing the Agent Governance Toolkit: Open-source runtime security for AI agents https://t.co/CHRV1iEWie

[Imran Siddique @mosiddi]

The Bottom Line:The role boundary is gone. You aren’t "just" an engineer or "just" a manager anymore. You are a Cognitive Systems Architect. The machine handles the labor. You handle the differentiation. If you aren't bringing your own ideas and your own hard-earned lessons to the table, you aren't using AI—you’re just watching it work.
Don’t manage agents. Architect their constraints.

#ScaleBySubtraction #AgentOS #SoftwareEngineering #AI #SystemArchitecture

[Imran Siddique @mosiddi]

The "Vibe Coding" Red Flag
I’m seeing a dangerous trend lately. Everyone is obsessed with "Vibe Coding"—throwing a vague spec at an LLM and hoping it builds the vision.

If you’re just handing over a spec and walking away, you aren't an architect. You’re a passenger.

In a world where the cost of "doing" is heading toward zero, being a passenger is a commodity. We need to stop trying to force old corporate titles into this new reality. I don’t care about being a "Manager of Agents" or a "Tech Lead of Bots." These aren’t humans. They don’t need a manager; they need Intent Injection.

Here’s what I’m actually seeing in the trenches:

[Imran Siddique @mosiddi]

3. Watching for "Agent Laziness"We used to measure PR velocity. Now, I’m watching for behavior. Is the agent getting lazy? When I ask it to analyze a meeting transcript, is it actually reading the data, or is it hallucinating an excuse to save itself the compute? Is it over-engineering because I didn't give it a "Scale by Subtraction" constraint?

[Imran Siddique @mosiddi]

@sanakhan122k thank you

[Imran Siddique @mosiddi]

Managing the SFI for one of Microsoft's most visited platforms has been very rewarding. We resolved existing challenges and strengthened future safety measures. By utilizing AI Transformation, we increased productivity by three times.

[Imran Siddique @mosiddi]

@grok Find key integration opportunities!

[Imran Siddique @mosiddi]

Multi-agent systems in prod look like microservices… but without an OS? Prompt safety won’t stop privilege escalation when Agent A hands off to Agent B. Time for a runtime supervisor / Agent Hypervisor.

Open-sourcing primitives: https://t.co/zE0kcuaDal

#AgenticAI #AgentOS #AISafety #DistributedSystems

https://t.co/1iSGF9wcY1