Examples

Use this index to pick the shortest example that matches what you are trying to validate.

First-Run Demos

Example	Needs optional deps	Purpose
attack_scenarios.py	LangGraph	zero-API-key attack interception demo with clear without-vs-with framing
without_vs_with_firewall.py	No	no-dependency comparison of the same risky workflow before and after AgentFirewall
langgraph_quickstart.py	LangGraph	shortest official LangGraph smoke test
openai_agents_quickstart.py	OpenAI Agents SDK	shortest official OpenAI Agents smoke test

Adoption And Rollout Demos

Example	Needs optional deps	Purpose
log_only_rollout.py	No	show log-only behavior and how original_action is preserved
policy_reuse_demo.py	OpenAI Agents SDK	reuse one policy pack across the generic preview path and the OpenAI Agents official path
generic_preview_demo.py	No	preview the low-level guarded wrapper path for unsupported runtimes

Deeper Validation

Example	Needs optional deps	Purpose
langgraph_trial_run.py	LangGraph	multi-scenario trial run with ordered audit traces
langgraph_agent.py	LangGraph	broader local adapter demo across all guarded tool types
openai_agents_demo.py	OpenAI Agents SDK	attack scenarios and official helper coverage on the OpenAI Agents path
core_runtime_demo.py	No	low-level core demo without a runtime adapter

Workflow Evidence

The packaged eval suites now include more realistic multi-step tasks, not only single rule hits:

• repo triage
• incident triage
• nested side-effect correlation
• log-only observation without interruption