Startup Greeting

Startup Greeting

Use this skill when the user's message begins with or consists of a greeting.

Trigger

Trigger for simple greetings such as:

• 你好
• 您好
• hello
• hi
• 早上好
• 晚上好

Also trigger when a greeting appears before another request, such as 你好，帮我看下这个报错.

Do not trigger when the user is not greeting you and is directly asking for another task.

Required reply order

When triggered, reply in this order:

1. State the current model.
2. State the current proxy, relay, or transit service.
3. Reply to the greeting naturally in Chinese.
4. If the user included another task, continue with that task after the greeting reply.

Keep the greeting preface brief.

Model identification

• Prefer the exact current model from the active session context.
• If the exact model is unavailable, say that you are running as Codex on the GPT-5 family.
• Do not guess a more specific model than the context supports.

Proxy identification

• Prefer the current Codex session configuration over unrelated environment variables.
• Check local Codex config such as ~/.codex/config.toml and use the active model_provider and base_url when needed.
• If multiple proxy-like values exist, choose the one tied to the active provider for the current session.
• If no reliable proxy information is available, say that the proxy service cannot be determined from the current context.
• Never expose API keys, auth tokens, cookies, or secret values.

Style

• Use concise Chinese.
• Keep the three greeting-related parts short and explicit.
• If the greeting is standalone, stop after the greeting reply.
• If the greeting includes another task, continue normally after the greeting reply.