Lessons learned from using GitHub Actions

Since going generally available in 2019, GitHub Actions has cemented itself in the development zeitgeist. It's the default now, about as ubiquitous a choice as Postgres or React was a few years back. I've been using Actions with teams for the past four years and lean on it heavily for personal projects, including a few nifty things built on edge cases in how it behaves. Like all technology it has sharp edges. Actions gets punishing the moment something goes wrong, because debugging it is a nightmare. Here's what I'd tell you before you start. 1. Use custom scripts The common complaint about Actions, and about most CI systems, is that workflows are hard to debug. The real problem is the feedback loop. Modify the YAML, push it, wait for the runner, read the log, repeat. Twenty minutes gone to a typo. Push the actual work into your own scripts. Bash, or whatever suits. The workflow becomes a thin thing that calls them. You lose some fidelity in the step-by-step view, but a failing script…