Common Patterns

Frequently used ACGP patterns and best practices.

---

Pattern: Pre-Execution Evaluation

Evaluate actions before execution:

result = steward.evaluate(trace)
if result.intervention in ["OK", "NUDGE"]:
    execute_action()

---

Pattern: Async Evaluation

For non-blocking governance:

async def governed_action():
    result = await steward.evaluate_async(trace)
    if result.intervention == "OK":
        await execute_action_async()

---

Pattern: Batch Evaluation

Evaluate multiple actions efficiently:

traces = [trace1, trace2, trace3]
results = steward.evaluate_batch(traces)

---

Pattern: Custom Interventions

Handle interventions based on your needs:

intervention_map = {
    "OK": lambda: execute(),
    "NUDGE": lambda: log_and_execute(),
    "FLAG": lambda: flag_for_review_and_execute(),
    "ESCALATE": lambda: request_human_approval(),
    "BLOCK": lambda: notify_and_abort(),
    "HALT": lambda: shutdown_agent()
}
intervention_map[result.intervention]()

---

More patterns coming soon

---

Implementation Guides