Commit like a librarian.

Commit like a librarian.

Coding agents leave working trees full of interleaved changes, and `git add -A` turns them into megacommits nobody can review, bisect, or blame. Three small shell scripts (`git addgrep`, `git restage`, `git post-fmt`) and two agent skills make focused, coherent commits the path of least resistance.
Claude Code has amnesia. So do PRs, changelogs, and your future self.

Claude Code has amnesia. So do PRs, changelogs, and your future self.

Technical Project Plans (TPPs) are living docs that brief four readers: the next session (when context runs out), the PR reviewer (who needs the why), the changelog writer at release time, and the engineer who inherits this code years from now. Works with Claude Code, Codex, and Gemini.
Claude picks the first idea that works. Make it pick the best one.

Claude picks the first idea that works. Make it pick the best one.

Coding agents plan by picking the first reasonable approach and running with it. Replan is a skill that forces iterative critique-and-refine cycles, surfacing better designs before a line of code is written. Works with Claude Code and Codex.
Most AI code reviews are noise. Here's how to fix that.

Most AI code reviews are noise. Here's how to fix that.

AI code review defaults to a wall of style nitpicks and speculative concerns. A skill that demands proof before reporting turns it into something worth reading. Works with Claude Code and Codex.
The LLM sycophancy antidote

The LLM sycophancy antidote

A dead-simple trick to make LLMs actually push back on your ideas instead of cheerfully agreeing with everything you say.
Serve AI-readable Markdown from Hugo

Serve AI-readable Markdown from Hugo

Hugo's custom output formats publish a clean Markdown version of every page and a /llms.txt index, with no HTML for LLMs to wade through.
Custom status lines for Claude Code

Custom status lines for Claude Code

Claude Code can display a persistent status line at the bottom of your terminal. Here's how to configure one that shows your working directory, model, effort level, and context usage.
How to VACUUM SQLite in WAL Mode (and why you need a checkpoint)

How to VACUUM SQLite in WAL Mode (and why you need a checkpoint)

In WAL mode, VACUUM doesn't shrink your database—it writes changes to the WAL file. Without a checkpoint afterward, you've done expensive work for nothing. Here's the proof and the fix.
Fix "Exit handler never called" with Node 22.5.0

Fix "Exit handler never called" with Node 22.5.0

Troubleshooting guide for resolving the "Exit handler never called" error in Node.js 22.5.0 that was breaking npm install operations in GitHub Actions builds.
Set up Wake-on-LAN with systemd

Set up Wake-on-LAN with systemd

Step-by-step guide to configuring Wake-on-LAN functionality with systemd on Linux systems to remotely power on suspended workstations and save energy.
Switch your cross-platform CI to GitHub Actions in 5 minutes

Switch your cross-platform CI to GitHub Actions in 5 minutes

Step-by-step guide to migrating your open source project's CI/CD from other providers to GitHub Actions with cross-platform testing.
If Something is Odd, Inappropriate, Confusing, or Boring, It Is Probably Important.

If Something is Odd, Inappropriate, Confusing, or Boring, It Is Probably Important.

A programming philosophy about paying attention to code that seems strange or confusing, as it often reveals important insights for future maintainability.
Uncertain, lazy, forgetful, & impatient: It’s what you want your code to be.

Uncertain, lazy, forgetful, & impatient: It’s what you want your code to be.

Explore programming patterns and tools that embrace uncertainty, laziness, forgetfulness, and impatience to create more robust and maintainable code.