Make Any Complex Topic Click With the Claude ELI5 Skill

ELI5 is a public Claude skill. It gives Claude instructions for explaining a hard topic in simpler language for a particular audience.

Open a terminal in a folder you don't mind using for this, then clone the repo and copy the skill folder into Claude Code's skills directory:

git clone https://github.com/DreambigOu/ELI5.git
cp -r ELI5/skills/eli5 ~/.claude/skills/eli5

Restart Claude Code, then run /skills to confirm ELI5 appears in the list.

For claude.ai, skip the terminal. Open the eli5 skill folder in the GitHub repo, download SKILL.md, and drag it into a chat. First turn on Code execution and file creation in Settings > Capabilities. An Enterprise admin can also turn on skill and plugin scanning in Organization settings. If that setting is off, read the file yourself before you trust it.

Pro tip: Skim SKILL.md before trusting any skill with real prompts. It's a plain-text file, so reading it takes a minute and tells you exactly what the skill will and won't do.

Try it once before doing anything else:

Explain like I'm five what skills are and how you can create one.

You can also point the skill at a document, codebase, or error message. The useful part is getting an explanation that fits the person reading it.

Trying a skill once tells you what it sounds like. A quick evaluation tells you whether it changes anything.

The ELI5 repo includes an evaluation setup. Ask Claude to use its format in eli5-workspace/evals.json:

Set up two test cases for me from this repo. You'll find the format in the GitHub instructions. Make one about the NBA playoffs and make one about golf.

Claude adds the prompts and pass/fail checks to evals.json. You do not have to build the evaluation system yourself.

Ask Claude for the command for each new test, then run it from the cloned repo folder. The command uses the test id in evals.json. In our example, NBA was test 4 and golf was test 5:

python3 eli5-workspace/run-evals.py --test=4
python3 eli5-workspace/run-evals.py --test=5

Keep --test=N. Without it, the script reruns every case in evals.json.

Each test answers the prompt with ELI5 and without it, then grades both against the checks and prints a pass-rate summary.

We tested NBA playoffs and golf, and both passed with ELI5 and without it. Those topics were easy enough that baseline Claude explained them well too. If that happens to you, pick a harder topic before deciding the skill is not helping.

Pro tip: If both versions pass, try something harder or more specific, such as a piece of your own code or a dense internal process. The ELI5 repo's published evals cover tougher tests, including explaining a database index to a 5-year-old, a codebase to a manager, and a git merge conflict to a 5th grader. That is the repo's earlier benchmark, not a result we reproduced in this guide.