fix(agent): remove reasoning and thinking from responses

[yetibot-mps]

This PR removes intermediate step-by-step reasoning and thinking from agent responses to keep them concise, direct, and playful.

• Updates the build-agent-prompt final instructions to forbid step-by-step narration, internal thinking, reasoning, or lists of justifications/explanations.
• Updates the settings.json configuration generated for the Gemini CLI to explicitly disable model reasoning/thinking with 'thinkingBudget': 0.