Your AI Agent Passed the Approval Check. Did the Side Effect?

Description and use

Agent Action Guard detects approval bypasses in OpenAI agents using static analysis and runtime witness to ensure real-world consequences are protected.

Problem

AI agents may bypass approval boundaries via alternate execution paths, leading to unintended real-world consequences.

Outcome

The combined static and runtime approach successfully identified bypass paths and confirmed boundary enforcement, enabling developers to validate safety boundaries in CI pipelines.

Workflow

• 1. Define approval boundary in agent workflow.
• 2. Run static scanner to detect possible counterexample paths.
• 3. Attach Runtime Witness hooks to OpenAI Agents SDK to record tool lifecycle events.
• 4. Execute agent with hooks in guarded mode to observe boundary usage.
• 5. Analyze runtime witness output for boundary verdict.
• 6. If OBSERVED_BYPASS or UNRESOLVED_TRACE, investigate and adjust agent or approval logic.

Evidence

The static scanner in the project looks for that second path.

Source

The runtime layer asks: «What path was actually observed when the protected consequence occurred?»

Source

If the execution evidence is incomplete, the system does not infer safety and does not infer a bypass.

Source