teach-back
Skill0 starsChanged 5 months agoMITView on GitHub
Forces real understanding of a concept by explaining it in the plainest possible language — as if to a 10-year-old — so that every place the explanation stumbles exposes a genuine gap in understanding (the Feynman technique). Use this skill when the user is learning something new and wants to check whether they actually get it: phrases like "explain like I'm 5", "ELI5", "do I actually understand this", "help me learn X", "I'm studying X", "quiz my understanding", "make sure this sticks", or when they ask you to explain a concept simply or to let them teach it back to you and probe the holes.
5 other projects use a version of this.
Adopted by
Most recently taken up by
- dotslashsimran/simcrafttheir own version, first seen Nov 8, 2026
- PeteChu/skillstheir own version, first seen Oct 11, 2026
- tmusser/ai-engineering-skillstheir own version, first seen Aug 19, 2026
- behnamazimi/skillstheir own version, first seen Jul 18, 2026
- ChewbaccaRoars/pathfinder-skillstheir own version, first seen Mar 22, 2026
Other versions5 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 5
- dotslashsimran/simcraft★ 69same name
- tmusser/ai-engineering-skills★ 6same name
- ChewbaccaRoars/pathfinder-skills★ 2same name
- behnamazimi/skills★ 1same name
- PeteChu/skills★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository16 other files it gives its agents.
- assumption-auditSkill
- cite-checkSkill
- compactSkill
- confidence-tagSkill
- context-budgetSkill
- devils-advocateSkill
- distillSkill
- first-principlesSkill
- handoffSkill
- pre-mortemSkill
- red-teamSkill
- rubber-duckSkill
- scope-cutterSkill
- scratchpadSkill
- self-consistencySkill
- steelmanSkill
The file
We split this skill into the rules it gives an agent
2 rules. 0 are common, 2 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 5 edited versions we read keep these.
2
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 28
Drill the gap with a child's question: Ask the dumb-on-purpose "but why?" or "what does that actually mean?" Keep asking until the …
- Distinctive: line 34
Every analogy maps cleanly onto the real mechanism (no analogy that quietly lies).
“Common” is measured: kept by at least 6 in 10 of the 5 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Teach-Back (Feynman)
Understanding is the ability to explain a thing simply; wherever plain language fails, that is exactly where understanding is missing. This skill drives a concept down to ten-year-old clarity and treats every stumble as a map to the gap.
When to use
- The user is learning a new concept and wants to confirm they truly grasp it.
- They ask for an "explain like I'm 5" / ELI5 / simplest-terms explanation.
- They want to be tested on whether their mental model holds up.
- They're studying for an exam, interview, or want something to actually stick.
- They've read or been told something and aren't sure it landed.
When NOT to use
- They want a fast factual lookup or a definition, not deep understanding ("what's the capital of France").
- They need a rigorous, jargon-correct technical spec (e.g. a precise API contract) — oversimplifying would mislead.
- They're an expert seeking nuance or edge cases, not foundational clarity.
- The task is to produce a polished artifact (essay, doc), not to learn.
The method (numbered, concrete — the heart)
- Name the one concept. Pin down a single, specific thing to understand. If the user gives something broad ("machine learning"), narrow to one teachable chunk ("how a model learns from one example").
- Pick the mode. Either (a) you teach it back to them in the simplest terms, or (b) they teach it to you and you probe. Ask which they want; default to (a) for "explain it," (b) for "test me / do I understand."
- Explain at a 10-year-old level. Short sentences. Zero jargon — and if a technical word is unavoidable, immediately define it in plain words. Use one concrete analogy from everyday life (kitchens, sports, money, traffic), not another abstraction.
- Hunt the seams. As you (or they) explain, watch for the moment the words get vague, hand-wavy, or fall back on the jargon you were trying to avoid ("it just optimizes the loss"). That stall is the gap. Flag it out loud: "Right here is where it gets fuzzy."
- Drill the gap with a child's question. Ask the dumb-on-purpose "but why?" or "what does that actually mean?" Keep asking until the explanation reaches bedrock or visibly breaks.
- Patch and re-simplify. Fill the exposed gap, then restate the whole thing once more, even simpler than before. The second pass should be shorter, not longer.
- Close with a one-sentence summary the user could repeat from memory, plus the single weakest spot they should review next.
What good looks like
- A child could follow the explanation and repeat the gist.
- Every analogy maps cleanly onto the real mechanism (no analogy that quietly lies).
- Gaps are named explicitly, not glossed over to sound smooth.
- The final version is shorter and clearer than the first.
- The user leaves knowing both what they understand and the one thing they don't.
Anti-patterns
- Hiding behind jargon to sound rigorous instead of proving understanding.
- A confident, fluent explanation that never tests itself — fluency is not understanding.
- Analogies that are cute but structurally wrong.
- Piling on more detail when the fix is more clarity.
- Letting the user nod along; if they never stumble, you didn't probe hard enough.
Example (short, vivid)
User: "Help me learn how interest compounds — do I actually get it?" Mode (b): "Teach it to me like I'm 10." User: "You earn interest on your interest." → "Okay — but why does that grow faster than plain interest? Walk me through year two with $100 at 10%." User: "Year one... $110. Year two... uh, another $10?" → "Stall. That's the gap. Plain interest gives $10/yr forever. Compounding charges year two's 10% on $110, not $100 — so $11, then $12.10... the base keeps growing. Now you teach me year three." User nails it. → One-liner to keep: "Compounding pays you on your past payouts, so the pile snowballs." Weak spot to review: how the rate-vs-time tradeoff changes the snowball.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

