|
| 1 | +# Agent Workflow System - README |
| 2 | + |
| 3 | +## Overview |
| 4 | + |
| 5 | +This directory contains **custom GitHub Copilot agent definitions** and workflow orchestration documentation for the agentic workflow test system. |
| 6 | + |
| 7 | +**Important**: These are custom agent definitions for use with GitHub Copilot's agent system, not standard GitHub @mentions. They must be invoked through GitHub Copilot, not directly in GitHub issues or pull requests. |
| 8 | + |
| 9 | +## Structure |
| 10 | + |
| 11 | +``` |
| 12 | +.github/agents/ |
| 13 | +├── develop-agent.md # Development agent definition |
| 14 | +├── test-agent.md # Testing agent definition |
| 15 | +├── document-agent.md # Documentation agent definition |
| 16 | +├── review-agent.md # Review agent definition |
| 17 | +├── workflow-orchestrator.md # Workflow chaining guide |
| 18 | +├── workflow-example.md # Complete example walkthrough |
| 19 | +└── README.md # This file |
| 20 | +``` |
| 21 | + |
| 22 | +## Quick Reference |
| 23 | + |
| 24 | +### Agent Sequence |
| 25 | +1. **develop-agent** → Implements code |
| 26 | +2. **test-agent** → Tests independently |
| 27 | +3. **document-agent** → Creates documentation |
| 28 | +4. **review-agent** → Reviews and approves/requests changes |
| 29 | + |
| 30 | +### Key Files |
| 31 | + |
| 32 | +- **Agent Definitions**: Individual `.md` files define each agent's role, responsibilities, and output format |
| 33 | +- **Workflow Orchestrator**: Explains how to chain agents together |
| 34 | +- **Workflow Example**: Shows a complete workflow example from start to finish |
| 35 | + |
| 36 | +## Using the Agents |
| 37 | + |
| 38 | +### Custom Agent Invocation |
| 39 | +These are **custom GitHub Copilot agents**. Invoke them through GitHub Copilot's custom agent system: |
| 40 | + |
| 41 | +``` |
| 42 | +Custom agent: develop-agent → completes task → provides context |
| 43 | +Custom agent: test-agent → completes task → provides context |
| 44 | +Custom agent: document-agent → completes task → provides context |
| 45 | +Custom agent: review-agent → approves or requests changes |
| 46 | +``` |
| 47 | + |
| 48 | +The `@agent-name` syntax in documentation is shorthand representing these custom agents, but they are invoked through GitHub Copilot, not as standard GitHub mentions. |
| 49 | + |
| 50 | +### Individual Agent Use |
| 51 | +Each agent can be invoked independently with specific instructions following the format in their definition file. |
| 52 | + |
| 53 | +### Workflow Chain |
| 54 | +Use the orchestrator guide to understand handoff protocols and context passing between agents. |
| 55 | + |
| 56 | +## Agent Characteristics |
| 57 | + |
| 58 | +### ✓ Independence |
| 59 | +Each agent operates independently without assuming context from previous agents. |
| 60 | + |
| 61 | +### ✓ Specialization |
| 62 | +Each agent is specialized for a specific role in the development workflow. |
| 63 | + |
| 64 | +### ✓ Clear Handoffs |
| 65 | +Each agent provides structured output for the next agent in the chain. |
| 66 | + |
| 67 | +### ✓ Iteration Support |
| 68 | +The workflow supports restart from develop-agent based on review feedback. |
| 69 | + |
| 70 | +## Current Status |
| 71 | + |
| 72 | +All agents are currently **placeholders** ready for concrete task definitions. The framework is in place for: |
| 73 | + |
| 74 | +- ✓ Agent definitions and roles |
| 75 | +- ✓ Workflow orchestration |
| 76 | +- ✓ Handoff protocols |
| 77 | +- ✓ Review feedback loop |
| 78 | +- ⏳ Specific development tasks (to be added) |
| 79 | +- ⏳ Testing frameworks (to be defined) |
| 80 | +- ⏳ Documentation standards (to be specified) |
| 81 | +- ⏳ Review criteria (to be established) |
| 82 | + |
| 83 | +## Extending the System |
| 84 | + |
| 85 | +### Adding New Agents |
| 86 | +1. Create a new `<agent-name>.md` file in this directory |
| 87 | +2. Follow the structure of existing agent definitions |
| 88 | +3. Update the workflow orchestrator to include the new agent |
| 89 | +4. Document handoff protocols |
| 90 | + |
| 91 | +### Customizing Existing Agents |
| 92 | +1. Edit the relevant agent `.md` file |
| 93 | +2. Maintain the standard output format |
| 94 | +3. Update the workflow example if needed |
| 95 | +4. Ensure compatibility with adjacent agents in the chain |
| 96 | + |
| 97 | +## Best Practices |
| 98 | + |
| 99 | +1. **Keep Agents Independent**: Each agent should work without relying on previous agent's assumptions |
| 100 | +2. **Clear Communication**: Use structured output formats for handoffs |
| 101 | +3. **Document Thoroughly**: Each agent should document what it did and what the next agent needs to know |
| 102 | +4. **Support Iteration**: Design for workflow restarts based on feedback |
| 103 | + |
| 104 | +## Further Reading |
| 105 | + |
| 106 | +- See `workflow-orchestrator.md` for detailed chaining instructions |
| 107 | +- See `workflow-example.md` for a complete example |
| 108 | +- See individual agent files for specific agent details |
0 commit comments