Senior engineering habits, forged into your AI agent.
Kiln is a set of 18 skills for AI coding agents. Each skill is a short instruction file the agent reads before a job, so it reviews code, writes tests, or ships a website the way a careful engineer would.
npx skills add KILN-skills/kiln
Free and MIT licensed. The installer supports Claude Code, Cursor, Codex, GitHub Copilot, Windsurf, Gemini CLI, and other agents that read SKILL.md files.
Grouped by the order work usually happens
You don't have to call a skill by name. Describe the task, like "review this diff" or "the site feels slow", and the agent loads the skill whose description matches. Click a stage to see its skills.
All 18 skills
Each row links to the skill's folder on GitHub, so you can read exactly what your agent will read.
What a skill actually looks like
A skill is one Markdown file, sometimes with a few helper scripts next to it. This is part of code-review: some lines skipped and long lines wrapped, wording unchanged.
- The description decides when it runs. The agent compares your request with this line, including phrases like "ada bug?" and "cek kode ini".
- The steps come in a fixed order. Find the scope, understand the intent, then check by severity.
- It runs your project's own checks. Type checker, linter, and tests, instead of guessing whether something is a bug.
Install
Run the command inside a project. The installer asks which skills and which agents you want, then copies the files into place.
- Pick from all 18
npx skills add KILN-skills/kiln - One skill only
npx skills add KILN-skills/kiln --skill code-review - Just list what's inside
npx skills add KILN-skills/kiln --list
Skills run with your agent's permissions, so read one before you install it. The installer also shows a security scan; you can see it on skills.sh.
Why this exists
These skills started as a personal folder we reached for every day while building with AI agents. Without them, the agent would rush: change code it hadn't read, skip the tests, call a slow page "optimized" without measuring it.
Writing the steps down fixed most of that. We put them in a public repo so you can use them, copy them, or make them better. If a skill gives you a bad result, open an issue with what you asked and what the agent did.