Hidde de Smet
Hidde de Smet

Meet me

Stories, journeys, and ideas for the creatively curious

Join me on a journey where every post brings a fresh perspective. This blog is for those who are driven by curiosity and a desire to learn.

Featured

98 skill folders, three coding setups, seven that don't exist: auditing your own SKILL.md library

I counted the skill folders across Claude Code, a shared agents catalog, and Copilot on my own machine. 98 total, including seven Google ADK Agents CLI folders pointing at nothing. Here's what a real audit finds and the checklist to run it yourself.

98 skill folders, three coding setups, seven that don't exist: auditing your own SKILL.md library

I counted the skill folders across Claude Code, a shared agents catalog, and Copilot on my own machine. 98 total, including seven Google ADK Agents CLI folders pointing at nothing. Here's what a real audit finds and the checklist to run it yourself.

The hidden costs of spec-driven development: when structure helps and when it slows you down

Spec-driven development fixes the 70% problem, but it also adds planning tax, review bottlenecks, and process overhead. Here is how to know when full specs are worth it and when to use a lighter path.

MCP finally got enterprise authorization: here's what changed

MCP's Enterprise-Managed Authorization extension just went stable. VS Code 1.123 shipped it in Preview with Entra ID support. One OAuth prompt per server is now a policy problem, not a user problem.

Terraform on Azure with guardrails: pre-commit, Trivy, and Anton Babenko's hooks

A small Azure Terraform setup wired to a pre-commit pipeline using Anton Babenko's pre-commit-terraform hooks: formatting, validation, linting, docs, and security scanning before code ever leaves your machine.

GitHub Copilot App vs VS Code Agents Window (2026): Which one should you use?

GitHub Copilot App is now open to Pro, Pro+, Max, Business, and Enterprise users. Compare it with the VS Code Agents Window and pick the right surface for each workflow.

AI coding agents need KPIs: how to measure speed, quality, reliability, and cost

A follow-up to the real cost of AI coding agents: how to turn usage-based billing, AI credits, model mix, and engineering outcomes into a practical KPI scorecard.