Agentic Workflow: An End-to-End Automated Software Development Pipeline Based on GitHub Issues

Key Takeaways: Agentic Workflow is an open-source framework that achieves full lifecycle automation from requirement analysis to acceptance testing, using GitHub Issues as the requirement entry point, GitHub Actions as the scheduling engine, and collaboration among multiple specialized AI Agents. Its design extends AI capabilities to the entire software engineering process, with clear responsibilities for each Agent and information transfer via deliverables, forming a clear and traceable pipeline.

Traditional CI/CD pipelines mainly focus on code integration and deployment, while Agentic Workflow aims to address the lack of automation in upstream software development stages (such as requirement analysis and architecture design), extending automation to the full lifecycle. Through collaboration among AI Agents, this project breaks through the limitations of single code generation and achieves an end-to-end automated process.

1. System Architecture: Divided into four layers—trigger entry (GitHub Issues), scheduling engine (GitHub Actions Workflow), AI Agent execution layer (each Agent has clear roles/inputs/outputs), and deliverable storage (independent directories in Git repositories).
2. Phased Process: Three phases—requirement analysis and QA (convert requirements into structured documents and review), architecture design/coding/test case development (generate architecture, code, test cases), acceptance testing (execute tests and publish reports).
3. Design Principles: Isolation (independent directories for Issues), explicit gates (manual review/CI checks), deliverable-driven (transparent progress), Agent-agnostic (easy replacement and testing).

1. Deployment Steps: Configure API Key → Create Issue → Trigger requirement analysis Workflow → Comment/approve → Automatically execute subsequent phases.
2. Technology Stack: Scheduling layer (GitHub Actions), AI execution layer (opencode CLI + Alibaba Cloud Baichuan API), storage (Git), language (Python testing framework).
3. Specific Implementation: Each phase has clear deliverables (e.g., requirement documents, architecture design, source code, test reports), with results published via PRs and Issue comments to ensure process traceability.

Agentic Workflow represents a new direction in software development automation, extending CI/CD to upstream stages and improving efficiency and quality. Its design leverages AI capabilities while maintaining process controllability and auditability through phase division, gate mechanisms, and deliverable-driven approaches, making it suitable for standardized application scenarios.

1. Prioritize use in application scenarios with high standardization to maximize automation benefits.
2. Follow design principles when expanding the system: maintain Agent-agnosticism (for easy replacement and upgrades), adopt explicit gate mechanisms (to ensure manual supervision), and be deliverable-driven (to transparentize the process).
3. Strictly configure the API Key according to steps during deployment to ensure all Workflows are triggered normally.