{
  "id": 9019599,
  "title": "From Epic to Merge: An End-to-End Workflow for Software Development with AI Agents",
  "url": "https://urgent.news/2026/09/21/from-epic-to-merge-an-end-to-end-workflow-for-software-development",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-21T21:55:40.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/rafael_dev/from-epic-to-merge-an-end-to-end-workflow-for-software-development-with-ai-agents-1ijn"
  },
  "original_language": "en",
  "account": "AI-powered coding agents are increasingly able to tackle individual software tasks. However, implementing a complete feature consisting of multiple interdependent tasks presents a more complex challenge. The article proposes an end-to-end workflow for transforming a software initiative into units of work that AI agents can execute safely, while minimizing human intervention.\n\nAn Epic serves as the source of intent and constraints for the software feature. It should include key elements such as the objective, problem context, scope, non-goals, requirements, tasks, acceptance criteria, dependencies, risks, and success metrics. The Epic provides a unified description of the system's purpose, but it is not the sole source of truth. The actual repository contains the technical reality of the system, including APIs, schemas, architectural decisions, infrastructure, tests, and implementation constraints.\n\nWhen assigning tasks to AI agents, it is crucial to consider task granularity. A task is considered sufficiently granular when it represents a coherent delivery, can be implemented and validated independently, produces a pull request understandable and testable without relying on undeclared changes, and has major architectural decisions already resolved. Task size should be measured by cohesion rather than lines of code or the number of files.\n\nThe proposed workflow begins with the Epic as the source of intent and constraints. Task dependencies are represented as a graph, with each task having clear input and output requirements. A planner agent is responsible for orchestrating the task execution, ensuring that tasks are completed in the correct order and dependencies are satisfied. The builder agent is then tasked with implementing each task, producing a pull request that can be reviewed, tested, and reverted independently.\n\nFinally, a reviewer agent validates the completed work, checking that the change is a single logical unit, can be understood and validated without relying on undeclared changes, and meets the specified acceptance criteria. Once approved, the Epic integration pull request is created, incorporating the changes into the main branch along with continuous integration testing.\n\nBy breaking down a feature into smaller, cohesive tasks and leveraging AI agents at each stage, the workflow minimizes human intervention while maintaining the necessary controls for risk management. This approach allows for efficient, safe, and reliable implementation of complex software features.",
  "summary": "Coding agents are becoming increasingly capable of implementing individual software tasks. Give an agent a repository, a clear issue, and enough context, and it can often inspect the codebase, modify files, write tests, and produce a working implementation. The harder problem starts one level above that. What happens when we need to implement an entire feature consisting of ten related tasks?…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}