| name | massgen-config-creator |
|---|---|
| description | Guide for creating properly structured YAML configuration files for MassGen. This skill should be used when agents need to create new configs for examples, case studies, testing, or demonstrating features. |
| license | MIT |
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-levelenable_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:
- Always read the authoritative docs first:
docs/source/development/writing_configs.rst - Read existing configs to understand current patterns
- Copy structure from similar configs, don't invent
- Test immediately after creating
- 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.