feat: Add Agent OS safety governance example#300
Conversation
This example demonstrates kernel-level safety for CrewAI agents: - Blocks dangerous file operations (rm -rf, shutil.rmtree) - Blocks privilege escalation (sudo, chmod 777) - Allows safe operations (file reads, analysis) - Full audit logging No dependencies required - standalone demo. Learn more: https://github.com/imran-siddique/agent-os
|
Hi team! 👋 Just checking in on this PR. I'd love to get feedback on the approach here. The example demonstrates how Agent OS can add a safety layer to CrewAI agents by blocking dangerous shell commands (rm -rf, sudo, etc.) at the kernel level rather than relying on prompt engineering. Happy to make any changes if needed. Let me know if you'd like me to:
Thanks for maintaining such a great examples repo! |
|
Friendly ping -- Agent OS was just merged into microsoft/agent-lightning (14k stars): microsoft/agent-lightning#478 -- Would love to get this example merged too! |
|
Update: Our AgentMesh trust layer was just merged into LlamaIndex (47k stars): run-llama/llama_index#20644. This is our second major integration merge this week after Microsoft's agent-lightning (14k stars). Would love to get this PR reviewed as well! |
|
Friendly follow-up! Since opening this PR, the Agent OS governance layer has been merged into three frameworks:
Would love to get this CrewAI safety example in front of the community too. Any feedback? |
|
Migration update: This project has officially moved to microsoft/agent-governance-toolkit under the Microsoft org. The code in this PR has been updated to reference the new location. Install via: All old personal repos ( |
|
Closing — this project has moved to microsoft/agent-governance-toolkit. Will re-submit fresh proposals from the Microsoft repo. Thank you! |
What this PR adds
A new example demonstrating kernel-level safety governance for CrewAI agents using Agent OS.
Why it matters
CrewAI agents can hallucinate dangerous operations like \
m -rf\, \\sudo\, and \\chmod 777\. Prompt engineering alone cannot reliably prevent these. Agent OS provides a kernel-level safety layer.
What the demo shows
How to run
\\�ash
cd agent_os_safety
python main.py
\\
No dependencies required - standalone demo.
Related