Does the World Need Another Agent Skills Repo?
Why I published the opinionated agent workflows I use for Jira, planning, code review, and reasoning-rich commit messages.
Luca Becker — Writing
Notes on AI-assisted development, infrastructure, and software engineering.
Latest posts
Six Months of Vibe Coding My Home Infrastructure
How AI turned a manual homelab into a multi-site GitOps setup, what broke along the way, and why businesses still need guardrails.
Renovate Can Update Almost Anything If You Teach It Regex
I finally discovered Renovate can update version strings in arbitrary files with its regex manager, and immediately used it for both my homelab and client infra.
Not All UUIDs Are Created Equal
Five UUID-looking identifiers, one subtle bug, and a reminder that platform ids do not always mean what you think they mean.
Inspecting My Balcony Battery's Cloud Telemetry
A tcpdump capture led to a closer look at the Marstek B2500-D cloud endpoint: plain HTTP, AES-128-ECB, and a static key.
Cleaning Up WIP Commits with Git Autosquash
A small git fixup and autosquash workflow for moving temporary WIP changes back into the commits where they actually belong.
Adding a Battery to My Balcony Solar and Reverse-Engineering Local Control
Slotting a Marstek B2500-D between my panels and inverter, expecting a cleaner local API, and ending up building two open-source Go services around its undocumented MQTT interface.
git fsck Saved My Draft. And I Didn't Know It Existed Till It Did.
My blog post disappeared from the working tree. git fsck --lost-found and git cat-file got it back. I'd never heard of either command before.
GitHub's Agentic Execution Environment
You could already run AI in GitHub Actions. gh-aw's real novelty is the sandboxed execution environment around it, and one Renovate review showed why that matters.
Hardening npm as a Consumer and Publisher
Axios got compromised, but the bigger lesson is how to harden npm as a consumer and, if relevant to you, as a publisher too.
What Happens When Reverse Engineering Gets Cheap
A benchmark on abandonware reverse engineering: one AI model refused, another reconstructed an old CD-key check in about 20 minutes.
The Biggest Overhaul to My Homelab Setup Yet
My homelab was already working fine, but I rebuilt it anyway to get easier rollbacks, simpler disaster recovery, and a faster way to ship changes with AI.
Stop Adding Secrets to Your mcp.json: Adding OAuth to Remote MCP Servers with Authelia
Add OAuth to remote MCP servers with an nginx gateway and Authelia. No more API keys in mcp.json. Just a URL and a browser login.
Aurora PostgreSQL Blue/Green Upgrade: 4 Gotchas
Four hard-won lessons from upgrading Aurora PostgreSQL 15 to 16 via blue/green deployment - from PostGIS compatibility to TypeORM hidden DDL statements.
Agent Skills: Teaching Your AI How to Actually Work
From babysitting commands to fire-and-forget confidence: how Agent Skills transformed my git workflow and why they matter for AI-assisted development.
Mining Ethereum Classic with Sunshine: Putting My Solar Surplus to Work
How I automated my RTX 3060 to mine Ethereum Classic only when my balcony solar panels export excess energy, with AI-assisted setup and Prometheus automation.
Level Up Agentic Coding with MCP #3: Build Your Own
When the MCP catalog doesn't have what you need, build it yourself. 40 minutes from idea to talking to Confluence.
The React Bug That Google Translate Causes
A mysterious production crash you can't reproduce locally. The culprit? Google Translate modifying the DOM behind React's back.
From Sunlight to Prometheus: My Balcony Power Plant Journey
How I set up a balcony solar system, built a Prometheus exporter with Cursor, and redesigned my Grafana dashboard using Claude Code with MCP servers.
Cursor 6 Months Later: From Tool to Thinking Partner
A retrospective on six months of AI-assisted coding with Cursor - how Planning Mode became a brainstorming partner, why code review is now the bottleneck, and what to expect in 2026.
AI Code Reviews: DIY vs Copilot Six Weeks Later
Six weeks running OpenCode and GitHub Copilot in parallel on production code. Real costs, actual bugs caught, and which approach wins when UX battles capability.
The State of Cursor, November 2025: When Sandboxing Leaks Your Secrets
Cursor's new sandbox security model can expose credentials from your home directory. How the switch from allow-lists to filesystem access created new security risks.
Cursor's New Features: One Hit, One Miss
Planning Mode proves Cursor can iterate thoughtfully, while Cursor Hooks feels rushed. A detailed review of both features from six months of daily use.
Level Up Agentic Coding with MCP #2: Stop Describing UI Issues
Add browser automation to your AI agent with Playwright MCP. See how visual inspection transforms UI debugging from 10+ iterations to 2-3.
aicommits v1.0.0: Six Months of AI-Assisted Development Come Full Circle
Announcing aicommits 1.0.0, a milestone release bringing flexible AI provider support, AWS Bedrock integration, and lessons learned from building AI tools with AI.
Trying Cursor's New CLI: More Than Just Desktop in Your Terminal
A hands-on review of Cursor's new CLI tool, covering installation quirks, model flexibility, GitHub Actions potential, and how it compares to desktop coding assistants.
Another Codex: OpenAI's Uncooperative Thinking Machine
Testing GPT-5-Codex against Claude Sonnet 4 and finding that long thinking times and uncooperative behavior kill developer productivity.
Level Up Agentic Coding with MCP #1: Web Search
Bridge the gap between RooCode and Cursor by adding web search capabilities using the DuckDuckGo MCP server for current information during development.
Can You Actually Code from Your Phone? Testing Cursor Agents
I spent a few evenings testing Cursor Agents to see if AI-powered development on isolated VMs could actually let you code on the go. Here's what I learned about the promises and limitations of autonomous coding agents.
Test Your SSL/TLS Configuration Before Your Users Do
Don't let certificate chain issues or outdated TLS configs bite you in production. Here's a bash tool that catches SSL/TLS problems before they catch you.
Agent Hooks: Amazon's Take on Reactive Development
Testing Kiro's Agent Hooks feature - a fascinating take on reactive development where your IDE automatically responds to changes. But does the execution match the innovative concept?
That Weird Scrolling Thing in Git Has a Name (And You Can Configure It)
Discover what pagers are, why git uses them, and explore modern alternatives to the default less pager for better command-line productivity.
First Look at Kiro: Amazon's New Agentic Coding Tool - Beta Impressions
Amazon's Kiro has brilliant architectural ideas but dangerous security flaws. My honest review after 4+ hours of testing - including why unpredictable command execution makes it too risky for real work yet.
Cursor vs VSCode Extensions: A Real-World Comparison of AI Coding Tools
After months of testing Cursor against free VSCode extensions like RooCode and KiloCode, here's an honest breakdown of what actually matters in day-to-day AI-assisted development.
RAG Isn't as Complex as You Think: Building a Zendesk Chatbot in 90 Minutes
How I went from thinking RAG required custom search APIs to building a functional knowledge base chatbot in one evening, using AI-assisted development to demystify vector embeddings.
AI Agent PoC Speedrun [Any%]: 90 Minutes to Working Demo
From weekend curiosity to client opportunity: How we built a working Postgres AI agent in 1.5 hours while chatting, and what it teaches us about the reality of AI-assisted development.
Building a Weather-Aware Lunch Bot with Cursor: When AI Development Actually Delivers
How I used Cursor to build a serverless Slack bot in just 2 hours of actual work - and why the hardest part wasn't the code, it was the corporate AWS discussions.
Building a Better AI Commit Tool: When the Original Just Isn't Cutting It
Sometimes the best solution is to fork it yourself. Here's how I rebuilt aicommits from the ground up to handle real-world consulting requirements - and why it took just two weeks with AI assistance.
Lovable.dev from a Senior Developer's Perspective: When AI Design Actually Works
A senior developer's honest review of Lovable.dev - exploring how AI-powered design tools are finally delivering on their promises and changing the development landscape.
Check Your TypeScript Packages Before You Ship Them
Don't let sneaky module resolution issues ruin your users' day. Here's a CLI tool that catches TypeScript package problems before you publish.
My AI Coding Journey: What Works, What Doesn't
Two months into AI-assisted development as a senior consultant - the key insights, best practices, and mindset shifts that transformed how I write code.
Welcome to My Blog: Sharing Insights on Modern Web Development
A brief introduction to my new blog where I'll share insights on modern web development, best practices, and the latest technologies I'm exploring.