100 ready-to-use agent team harnesses for everyday life and work, built for Claude Code.

Each harness orchestrates 4-5 domain-expert agents collaborating via SendMessage in production-grade workflows.

Copy any harness folder's .claude/ directory into your project:

# Example: apply the youtube-production harness to your project
cp -r 01-youtube-production/.claude/ /path/to/my-project/.claude/

Each harness folder contains a CLAUDE.md with structure and usage details.

Every harness follows a consistent structure:

{NN}-{harness-name}/
└── .claude/
    ├── CLAUDE.md                          # Project overview
    ├── agents/
    │   ├── {agent-1}.md                   # Specialist agent definition
    │   ├── {agent-2}.md
    │   ├── {agent-3}.md
    │   ├── {agent-4}.md
    │   └── {agent-5}.md (optional)
    └── skills/
        ├── {orchestrator}/
        │   └── skill.md                   # Orchestrator skill
        ├── {agent-skill-1}/
        │   └── skill.md                   # Agent-extending skill
        └── {agent-skill-2}/
            └── skill.md                   # Agent-extending skill

Every harness meets these criteria:

• Agent Team Mode — Direct communication via SendMessage, cross-validation
• Domain Expertise — Real-world frameworks and methodologies embedded
• Output Templates — Structured output formats per agent
• Dependency Management — Task ordering with parallel execution
• Error Handling — Fallback strategies for failure scenarios
• Scale-Based Modes — Full / reduced / single-agent modes
• Test Scenarios — Normal / existing-file / error flows (3 types)
• Trigger Boundaries — Should-trigger + NOT-trigger defined
• Agent-Extending Skills — 2-3 domain-specific skills that amplify agent capabilities

Each harness embeds real-world frameworks and methodologies:

Apache License 2.0 — See LICENSE