The Agentic Coding Series

This nine-part series documents how I replaced ad-hoc "vibe coding" with a structured team of AI agents – from planning through implementation to end-to-end testing – built with Claude Code and GitHub Copilot.

Each part covers one concrete step of the workflow, including the real handoffs, checkpoints, and the mistakes I had to correct along the way. Start with part one, or jump directly to the topic that interests you.

Start with Part 1
  1. 1

    The End of the "One-Prompt Wonder" – Why We Need Agent Teams

    Why simple prompting and "vibe coding" fail on complex projects – and the four systemic problems that forced me to rethink my entire development workflow.

  2. 2

    The Strategists – Business Analyst & Solution Architect

    How two specialized AI agents steer the planning process in the right direction through Human-in-the-Loop checkpoints – and why that matters more than pure speed gains.

  3. 3

    The Makers – Implementation Plan & Developer

    How an atomic battle plan and a focused Developer Agent close the gap between concept and code – and why Human-in-the-Loop works differently here than in the planning phase.

  4. 4

    The Verifier – The E2E Test Agent

    Why a green build doesn't mean a feature works – and how a fifth, specialized agent turns the use case into an executable end-to-end test, closing the loop back to the original requirement.

  5. 5

    The Workflow in Action – From Vague Wish to Finished Commit

    A concrete walkthrough through all five phases of the agent team – from the first vague feature wish to a tested commit. Including the real handoffs, the checkpoints, and the moment where I had to correct the agents.

  6. 6

    Creating Agents – Templates as Shortcut or Scaffolding from Scratch

    Two ways to your own agent team in Claude Code and GitHub Copilot – adapt ready-made templates or build from scratch with /agents or /create-agent. Plus the trick for translating a Claude template into a Copilot-compliant variant via GitHub Copilot.

  7. 7

    End-to-End Development with Agents and MCP

    How I integrate Figma and Appwrite MCPs into my agent workflow to cleanly connect frontend and backend without manual translation.

  8. 8

    How Do I Measure the Positive Effect of Agents? My Metrics for Agent Performance

    Which metrics can be used to measure the positive effect of AI agents in the development process? My approaches to agent performance measurement.

  9. 9

    The Dark Side of AI Agents – Skill Atrophy, Cognitive Debt, and the Way Back to Control

    Which risks can arise from using AI agents in the development process? An honest look at skill atrophy, cognitive debt, hidden costs – and why the problem is real, but perhaps temporary.