[Bug]: Initial “team collaboration” instruction is ignored until retried after interruption

## CoPaw Version

v1.0.1

## Description

When I explicitly ask CoPaw to complete a task in team collaboration mode, the first request does not actually trigger multi-agent collaboration.

Instead, it behaves like a normal single-agent execution. Only after I interrupt the output and send another follow-up instruction asking it again to work in team collaboration mode does it start a multi-agent collaboration flow.

This is problematic because the original instruction is already explicit, so the system should honor it on the first attempt rather than only after an interruption and retry.

**Related PR(s):** #N/A

**Security considerations:** None known.

## Component(s) Affected

- [x] Core / Backend (app, agents, config, providers, utils, local_models)
- [x] Console (frontend web UI)
- [ ] Channels (DingTalk, Feishu, QQ, Discord, iMessage, etc.)
- [x] Skills
- [ ] CLI
- [ ] Documentation (website)
- [ ] Tests
- [ ] CI/CD
- [ ] Scripts / Deploy

## Environment

- **CoPaw version:** v1.0.1
- **OS:** latest macOS
- **Install method:** curl -fsSL https://copaw.agentscope.io/install.sh | bash
- **Python version (if applicable):** Python 3.12

## Steps to Reproduce

1. Start CoPaw with multiple agents available.
2. Send a prompt such as:

`请以团队协作的形式完成以下任务：调查当前用于 Continual Learning 的 benchmark ARC-AGI-3 是否是一个比较好的 benchmark。`

3. Observe how CoPaw starts processing the request.

## Actual vs Expected

- **Actual:**
The first request does not trigger a real multi-agent collaboration flow, even though the instruction explicitly says to complete the task in team collaboration mode.

- **Expected:**
CoPaw should immediately start a multi-agent collaboration workflow on the first request, without requiring interruption and retry.

## Logs / Screenshots