SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsmixseek-team-config

mixseek-team-config

Agents & models

View repository

Updated 10-15 21:13

About this skill

MixSeek Team Configuration Generation

Overview

Generates the MixSeek-Core team configuration file (team.toml). It defines the structure of the Leader Agent and Member Agents, the models to use, system instructions, and so on.

Prerequisites

  • The workspace must be initialized (seemixseek-workspace-init)
  • The environment variable MIXSEEK_WORKSPACE must be set (recommended)

Usage

Step 1: Requirements Hearing

Confirm the following with the user:

  1. Team purpose: What the team should achieve
  2. Required Member Agents: What roles of agents are needed
  3. Model to use: Whether to use the default (Gemini) or a specific model

Step 2: Propose Team Composition

Based on the user's requirements, propose the following:

  • team_id: A unique ID (kebab-case recommended)
  • team_name: Display name
  • Leader Agent: Instructions and model
  • Member Agents: Role, type, and model of each agent

Step 3: Generate the Configuration File

Generate the configuration file based on the following template:

[team]
team_id = "team-id"
team_name = "チーム名"
max_concurrent_members = 15

[team.leader]
system_instruction = """
あなたはチームのリーダーです。

## ツール呼び出しルール
| ツール名 | 用途 |
|---------|------|
| do_task | [役割]を実行 |

do_taskツールを使用してタスクを実行してください。
"""
model = "google-gla:gemini-2.5-pro"
temperature = 0.7
timeout_seconds = 300

[[team.members]]
agent_name = "member-name"      # 内部識別子（DB記録用）
agent_type = "plain"
tool_name = "do_task"           # ★ Leaderが呼び出す名前（system_instructionで使用）
tool_description = "このエージェントは[役割]を担当します"
model = "google-gla:gemini-2.5-flash"
system_instruction = """
あなたは[役割]を担当するエージェントです。
[詳細な指示]
"""
temperature = 0.2

agent_name vs tool_name (Important)

  • agent_name: Internal identifier (for logs and DB records)
  • tool_name: The tool name that the Leader calls(used in system_instruction)

In the Leader's system_instruction, always usetool_name.

Step 4: Save the File

Save the generated configuration to the following path:

$MIXSEEK_WORKSPACE/configs/agents/team-<team-id>.toml

Step 5: Validate the Configuration File (Required)

Be sure to run validation after generation.

uv run python skills/mixseek-config-validate/scripts/validate-config.py \
    $MIXSEEK_WORKSPACE/configs/agents/team-<team-id>.toml --type team

If validation succeeds, report the result to the user. If it fails, check the error details and fix the configuration.

Member Agent type

Type Description Use case
plain Basic text generation General analysis, summarization, and generation tasks
web_search With web search capability Retrieving the latest information, research tasks
code_execution With code execution capability Calculation, data processing, script execution
web_fetch With web fetch capability Retrieving data from specific URLs
custom Custom plugin Integration of custom tools

For details, see references/MEMBER-TYPES.md .

Recommended Models

Use case Model Reason
Leader Agent google-gla:gemini-2.5-pro High-quality instruction understanding and coordination
Member Agent (standard) google-gla:gemini-2.5-flash Fast response
Member Agent (high quality) google-gla:gemini-2.5-pro For complex tasks
Code execution anthropic:claude-sonnet-4-5-20250929 code_execution support

Example

Creating a Web Search Team

User: Web検索と分析ができるチームを作って

Agent: チーム設定を提案します:

       チームID: team-web-research
       チーム名: Web Research Team

       構成:
       - Leader Agent: gemini-2.5-pro（チーム調整）
       - web_researcher (web_search): Web検索担当
       - analyst (plain): 情報分析担当

       この構成でよろしいですか？

User: はい

Agent: 設定ファイルを生成しました:
       configs/agents/team-web-research.toml

       検証を実行します...
       ✅ 検証成功: 設定ファイルは有効です。

       実行コマンド:
       mixseek team "調べたいこと" --config configs/agents/team-web-research.toml

Example generated configuration file

[team]
team_id = "team-web-research"
team_name = "Web Research Team"
max_concurrent_members = 15

[team.leader]
system_instruction = """
あなたはWeb調査チームのリーダーです。

ユーザーからの質問に対して、以下の手順で調査を進めてください:
1. search_webツールで関連情報のWeb検索を実行
2. 取得した情報をanalyze_dataツールで分析
3. 分析結果を統合して、ユーザーに回答を提供

## ツール呼び出しルール
| ツール名 | 用途 |
|---------|------|
| search_web | Web検索を実行 |
| analyze_data | 情報を分析 |

各ツールを適切に使い分けてください。
"""
model = "google-gla:gemini-2.5-pro"
temperature = 0.7
timeout_seconds = 300

[[team.members]]
agent_name = "web_researcher"   # 内部識別子
agent_type = "web_search"
tool_name = "search_web"        # ★ Leaderが呼び出す名前
tool_description = "Web検索を実行し、最新の情報を取得します"
model = "google-gla:gemini-2.5-flash"
system_instruction = """
あなたはWeb検索の専門家です。
与えられたトピックについて、信頼性の高い情報源からの最新情報を検索してください。
検索結果には出典URLを含めてください。
"""
temperature = 0.2

[[team.members]]
agent_name = "analyst"          # 内部識別子
agent_type = "plain"
tool_name = "analyze_data"      # ★ Leaderが呼び出す名前
tool_description = "収集した情報を分析し、洞察を提供します"
model = "google-gla:gemini-2.5-flash"
system_instruction = """
あなたは情報分析の専門家です。
提供された情報を分析し、以下を含むレポートを作成してください:
- 主要なポイントの要約
- 情報の信頼性評価
- 結論と推奨事項
"""
temperature = 0.2

Troubleshooting

An error occurs when running the team

  1. Configuration file validation: mixseek-config-validateValidate with a skill
  2. API key check: Check whether the API key corresponding to the model in use is set
  3. Workspace path: MIXSEEK_WORKSPACECheck whether is set correctly

Member Agent count limit

max_concurrent_membersThe range for is 1-50. The default is 15. If you use many agents, adjust this value.

References

  • TOML schema details: references/TOML-SCHEMA.md
  • Member Agent type: references/MEMBER-TYPES.md
  • Available models: skills/mixseek-model-list/references/VALID-MODELS.md