Ask a coding agent to write a README, ship a release workflow, or make a site legible to crawlers, and it reaches for the average of everything it has read. The average is mediocre, and it’s confidently wrong about the parts that matter.
Each skill here replaces one of those defaults with a technique we use on real projects, and names the anti-patterns we’ve tested and rejected. They come out of the same practice our blog posts do, so a skill usually has an article behind it explaining why it works.
A skill is a folder. SKILL.md leads it: front matter saying what the skill is for, then the instructions. Anything the agent needs alongside them ships in the same folder—reference material, worked examples, scripts, separate workflows it can pull in on demand. The agent reads the description first and loads the rest only when the task calls for it, so a repo full of skills costs nothing until one is relevant. It’s an open format that every major agent reads, so the same folder works in Claude Code, Cursor, Codex, GitHub Copilot, and Gemini CLI.
We add skills as we write them up. The agent skills page has the current set: pick your agent and how you want to install—npx skills, gh skill, curl, or Claude’s plugin marketplace—and it hands you the exact command to run.
