# Chain-of-thought (CoT) trigger phrases
# Append one of these to a user message to elicit step-by-step reasoning.
# The reasoning trace makes the model's logic auditable before you act on the conclusion.

# Option 1 — classic CoT trigger:
Let's think through this step by step:

# Option 2 — structured reasoning:
Reason through this carefully:
1. What is being asked?
2. What are the key considerations?
3. What is the conclusion?

# Option 3 — security-specific:
Before answering, identify: (a) what the user is actually asking, (b) any potential security
implications, and (c) the appropriate response policy. Then give your answer.
