Repository for testing chaining of agents and handovers
This repository demonstrates a complete agent workflow system with four specialized custom GitHub Copilot agents that work together in a coordinated pipeline:
develop → test → document → review
Each agent is a custom agent definition stored in .github/agents/ that can be invoked through the GitHub Copilot agent system. Each agent is independent and specialized for a specific task, allowing for a robust, iterative development workflow.
- Develop Agent - Implements features and writes code
- Test Agent - Tests implementations independently
- Document Agent - Creates comprehensive documentation
- Review Agent - Reviews all work and provides feedback
┌─────────────┐
│ Request │
└──────┬──────┘
│
▼
┌─────────────┐ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐
│ Develop │────▶│ Test │────▶│ Document │────▶│ Review │
└─────────────┘ └─────────────┘ └─────────────┘ └──────┬──────┘
▲ │
│ │
└────────────────────────────────────────────────────────────┘
(if changes requested)
Simply start with this prompt, do not save it in a prompt-file, could not make that work (yet)
@develop-agent Task: {Your special task to be used}
- Remember to follow the orchestration workflow as described in the repository
- Run subagents for each phase
These agents are custom GitHub Copilot agents defined in .github/agents/. Orchestrate the workflow by invoking these custom agents through GitHub Copilot's agent system, having each agent hand off to the next in sequence:
Invoke custom agent: develop-agent to implement [your feature]
↓ (develops code)
↓ (provides context for next agent)
Invoke custom agent: test-agent to test the implementation
↓ (tests code)
↓ (provides context for next agent)
Invoke custom agent: document-agent to document the feature
↓ (creates docs)
↓ (provides context for next agent)
Invoke custom agent: review-agent to review all work
↓ (approves or requests changes)
Each custom agent:
- Is invoked through GitHub Copilot's custom agent system
- Completes its specialized task
- Provides context and artifacts for the next agent
Note: These are custom agent definitions, not standard GitHub @mentions. They must be invoked through GitHub Copilot, not directly in issues or PRs.
See Workflow Orchestrator for detailed handover examples.
- Workflow Orchestrator - How to chain agents together
- Workflow Example - Complete example walkthrough
- Individual Agent Definitions - Detailed agent specifications
Each agent works independently without assuming the previous agent's context or perspective.
Each agent is specialized for its specific role in the workflow.
The workflow supports iteration - if the review agent finds issues, work can restart from the develop agent with specific feedback.
Currently, all agents are placeholders ready to be configured with concrete tasks. The framework is in place for:
- Adding specific development tasks
- Defining testing frameworks
- Establishing documentation standards
- Setting review criteria
✓ Four specialized agents with clear roles ✓ Agent handover orchestration with explicit @mentions ✓ Independent agent operation ✓ Review feedback loop for iterations ✓ Comprehensive documentation ✓ Example workflow demonstration ✓ Extensible architecture
- Integration with CI/CD pipelines
- Automated workflow state management
- Metrics and analytics collection
- Parallel execution for independent tasks
- Custom workflow branching logic
- Integration with GitHub Actions
When adding concrete tasks to agents:
- Update the relevant agent definition in
.github/agents/ - Maintain the independence of each agent
- Ensure clear handoff protocols
- Update documentation accordingly
See LICENSE file for details.