Open Source

Hamburg / Remote

Agentic Engineering Harness as an Open-Source Project

Initiator and maintainer

July 2026

2 months

Self-employed

Project work

Hamburg / Remote

Relevance

Why this case matters

The case shows how problem understanding, responsibility, decisions, and implementation came together in this context.

Impact

Built a publicly installable Skills and harness system for evidence-backed autonomy across product clarification, development, delivery, operations, and learning.

What carries forward today

The repository is the public, installable part of my Agentic Operating Model: agents work across the product and delivery flow with versioned rules, Skills, checks, review loops, and explicit responsibility boundaries.

Open sourceAgentic engineeringMaximum evidence-backed autonomyMulti-repository harnessDDD + TDDAgile testingDevSecOpsReview loops

Defensible proof

MIT

Publicly reusable

Source, methods, and Skills are inspectable and installable on GitHub.

27 Skills

Launch catalog

Standalone product, engineering, harness, and craft Skills at public launch.

4 clients

Installation smoke test

Codex, Cursor, Claude Code, and Gemini CLI are verified in the repository.

Where this experience creates value

For developers and teams that want to trust agents with more than isolated coding prompts.
For product organizations that connect high autonomy with DDD, TDD, agile testing, DevSecOps, operations, and inspectable learning.

Case context

Overview

The Agentic Engineering Harness is not a prompt pack. It is my public working
model for using agents reliably across the software product flow: from signal
and triage through domain understanding, prototypes, implementation, and review
to release, operations, outcome review, and the next learning step.

The central goal is maximum evidence-backed autonomy. For understood work,
agents carry the loop through tests, quality boundaries, independent review,
commit, and authorized push. People remain responsible for strategy, investment
boundaries, new domain semantics, high risks, and genuine authority boundaries.

Responsibility

Activities

Standalone Skills for product craft, frontend craft, backend craft, DDD, TDD, testing, compliance, harness design, and agent work
Harness templates for single repositories, multi-repository systems, and multiple teams
Versioned infrastructure desired state and executable compliance rules: IaC plan/apply, GitOps-near reconciliation, drift signals, and policy as code with clear human accountability
A product and delivery model built around value streams, small vertical slices, disposable prototypes, and production learning
Rules for Git-owned evidence, current toolchains with cooldown, review loops, rollback, observability, and agent stewardship
Thin adapters for different agent hosts without hiding their differences behind false uniformity

Operating mode

Methodology

Many agent setups transfer old planning and approval logic onto faster tools:
first a complete specification, then implementation, then tests, then handover.
The harness follows how product and engineering work actually evolve: shared
understanding, the cheapest decisive experiment, vertical TDD slices, real
feedback, and versioned learning.

The repository deliberately remains an evolving, inspectable open-source
project. It is not a finished promise of an autonomous software factory and not
a retrospective AI label for older work.

Technical context

Technology stack

The tools are not the point by themselves. What matters is which system layers had to work together.

5Areas

18Technologies

Agentic EngineeringAgentic Operating ModelDomain-Driven DesignTest-Driven DevelopmentAgile TestingProduct EngineeringEnd-to-End Delivery

Multi-Agent Pipelines

GitHubCodexClaude CodeCursorGemini CLI

GitHub ActionsCI/CD PipelinesRenovate

DevSecOpsObservability

Next step

If you need similar responsibility, we can identify the next useful leverage point directly.

Send the situation, goal, and decision in front of you. I will respond personally with a clear view of the value I could add.

Discuss the next leverage point