feat: add AI Agent Governance with Aegis

[Acacian]

Summary

• Adds a Streamlit app demonstrating policy-based governance for AI agent tool calls using Aegis
• YAML policies control approval gates (auto, approve, block) with risk classification and full audit logging
• No API keys required — the demo uses a simulated agent executor so reviewers can run it immediately

What is Aegis?

Aegis (pip install agent-aegis) adds governance to AI agents in one line. Define YAML policies that intercept tool calls, classify risk, enforce approval gates, and log every decision. Works with LangChain, CrewAI, OpenAI Agents SDK, Anthropic, and MCP.

Files Added

File	Description
advanced_ai_agents/single_agent_apps/ai_agent_governance_aegis/aegis_governance_app.py	Streamlit app with 3 agent scenarios and editable YAML policy
advanced_ai_agents/single_agent_apps/ai_agent_governance_aegis/README.md	Setup instructions and explanation
advanced_ai_agents/single_agent_apps/ai_agent_governance_aegis/requirements.txt	Dependencies (agent-aegis, streamlit)
README.md	Added entry under Advanced AI Agents

How to Run

cd advanced_ai_agents/single_agent_apps/ai_agent_governance_aegis
pip install -r requirements.txt
streamlit run aegis_governance_app.py

Test Plan

• pip install -r requirements.txt installs cleanly
• streamlit run aegis_governance_app.py launches without errors
• All three scenarios (CRM, Email, Data) produce correct results
• Editing the YAML policy changes agent behavior in real time
• Delete actions are blocked by policy and show in audit trail

[awesomekoder]

Thanks for the submission! The governance concept is interesting, but the code uses a SimulatedAgentExecutor with no actual LLM calls. For this repo (awesome-llm-apps), we need tutorials that use an LLM/AI model. If you can build a version where a real LLM agent is governed by Aegis policies, that would be a stronger fit.