mcpField
M
massgen-config-creator

massgen-config-creator

github.com/massgenAgent workflows1.1k90 installsUpdated 20w ago
Repository

About massgen-config-creator

massgen-config-creator is a skill that guides the creation of YAML configuration files for MassGen agents. It focuses on following existing conventions, consulting the authoritative guide, and validating configurations.

Key features

  1. 01Guidance for MassGen YAML configs
  2. 02Convention and property placement reminders
  3. 03Examples for single- and multi-agent configs
  4. 04Config testing and validation workflow

Use cases

  1. 01Create example configs for new features
  2. 02Prepare reusable multi-agent workflows
  3. 03Test backend or tool integrations
  4. 04Share consistent configuration patterns

MCP support

The skill references MassGen's MCP integration documentation and identifies the MCP configuration category. It does not itself provide MCP server functionality.

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/massgen/massgen --skill massgen-config-creator

After installing, type/massgen-config-creatorin Claude Code to confirm it loaded.

Details

From the project README

Config Creator

This skill provides guidance for creating new YAML configuration files that follow MassGen conventions and best practices.

Purpose

The config-creator skill helps you create well-structured, validated configuration files for MassGen agents. It ensures consistency across the codebase and helps avoid common mistakes.

When to Use This Skill

Use the config-creator skill when you need to:

  • Create example configs demonstrating new features
  • Write configs for case studies or releases
  • Build reusable multi-agent workflow configs
  • Test new backend or tool integrations
  • Share configuration patterns with users

Authoritative Documentation

IMPORTANT: The primary source of truth for config creation is:

📖 docs/source/development/writing_configs.rst

This file contains:

  • Complete config creation workflow
  • All current conventions and rules
  • Property placement reference
  • Validation checklist
  • Common patterns and examples
  • Up-to-date templates

Always consult this document for the latest configuration standards.

Critical Rules (Quick Reference)

1. Never Invent Properties

ALWAYS read 2-3 existing configs first to understand current conventions:

# Find similar configs
ls massgen/configs/tools/{category}/

# Read examples
cat massgen/configs/basic/multi/two_agents_gemini.yaml
cat massgen/configs/tools/mcp/filesystem_claude.yaml
2. Property Placement Matters
  • cwd → BACKEND-level (individual agent workspace)
  • context_paths → ORCHESTRATOR-level (shared read-only files)
  • enable_web_search → BACKEND-level
  • enable_planning_mode → ORCHESTRATOR.COORDINATION-level

See docs/source/development/writing_configs.rst for complete property reference.

3. Key Conventions

✅ DO:

  • Prefer cost-effective models (gpt-5-nano, gpt-5-mini, gemini-2.5-flash)
  • Give all agents identical system_message
  • Use separate workspaces per agent
  • Include "What happens" comments explaining execution flow

❌ DON'T:

  • Reference massgen v1 or legacy paths
  • Invent new properties
  • Suggest cleanup commands that delete logs

Quick Start Workflow

Step 1: Research Existing Configs
# Find configs in your category
ls massgen/configs/tools/{relevant_category}/

# Read 2-3 similar examples
cat massgen/configs/basic/multi/two_agents_gemini.yaml
Step 2: Copy and Adapt
  • Copy a similar config as your starting point
  • Adapt values, never invent properties
  • Follow the structure from existing configs
Step 3: Test
massgen --config massgen/configs/tools/{category}/{your_config}.yaml "Test prompt"
Step 4: Validate

Refer to the validation checklist in docs/source/development/writing_configs.rst

File Naming and Location

Naming Pattern:

{agent_description}_{feature}.yaml

Location Categories:

  • massgen/configs/basic/ - Simple examples
  • massgen/configs/tools/filesystem/ - Filesystem operations
  • massgen/configs/tools/web-search/ - Web search
  • massgen/configs/tools/code-execution/ - Code execution
  • massgen/configs/tools/multimodal/ - Image, vision, audio
  • massgen/configs/tools/mcp/ - MCP integrations
  • massgen/configs/tools/planning/ - Planning mode

Common Patterns (Quick Reference)

Single Agent
agent:  # Singular
  id: "my_agent"
  backend:
    type: "claude"
    model: "claude-sonnet-4"
Multi-Agent
agents:  # Plural
  - id: "agent_a"
    backend:
      type: "openai"
      model: "gpt-5-mini"
    system_message: "Shared task description"

  - id: "agent_b"
    backend:
      type: "gemini"
      model: "gemini-2.5-flash"
    system_message: "Shared task description"
With Filesystem Access
agents:
  - backend:
      cwd: "workspace1"  # Backend-level

orchestrator:
  context_paths:  # Orchestrator-level
    - path: "massgen/configs/resources/v0.0.29-example/source"
      permission: "read"

Reference Files

Primary Documentation:

  • Config writing guide: docs/source/development/writing_configs.rst ⭐ START HERE
  • YAML schema reference: docs/source/reference/yaml_schema.rst
  • Example configs: massgen/configs/

Supporting Documentation:

  • Supported models: docs/source/reference/supported_models.rst
  • Backend configuration: docs/source/user_guide/backends.rst
  • MCP integration: docs/source/user_guide/mcp_integration.rst

Tips for Agents

When creating configs programmatically:

  1. Always read the authoritative docs first: docs/source/development/writing_configs.rst
  2. Read existing configs to understand current patterns
  3. Copy structure from similar configs, don't invent
  4. Test immediately after creating
  5. When in doubt, consult the full guide in docs/source/development/writing_configs.rst

This skill is a quick reference guide. For comprehensive, up-to-date information, always refer to the official documentation files listed above.

Comments

Markdown supported