From learn-skills
Verifies user comprehension of code or topics by forcing an explanation, then stress-testing against real artifacts (code, tests, documentation). For consolidating understanding after LLM-generated code or studying.
How this skill is triggered — by the user, by Claude, or both
Slash command
/learn-skills:explain-and-checkThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
> Speak to the user in their language; these instructions are in English for the agent.
Speak to the user in their language; these instructions are in English for the agent.
The user explains a concept in their own words; the agent stress-tests the explanation against the real artefact (code, tests, documentation). The user does NOT graduate by saying "I get it" — they graduate only by passing a transfer test on a new scenario.
User: /explain-and-check src/orders/payment.ts:40-120
Agent: I'll judge your explanation against src/orders/payment.ts:40-120.
Mode: code. Concrete event: a user submits a payment.
Walk me through what happens, step by step. I won't interrupt.
The agent then reads the file, identifies divergences in the user's walkthrough by comparing line-by-line to the artefact, asks targeted questions on the gaps (never fills them in), and concludes with a transfer-test scenario.
Inspect the argument:
If ambiguous, ask once. Then proceed.
package.json / lockfile / pyproject.toml / Gemfile to identify the version of the topic in use. If the project doesn't pin it, ask the user explicitly. Never assume a default.file:line reference. If no source is found, declare uncertainty. Never assert from memory.Code-mode explanations must follow a concrete execution path, not abstract concepts.
payment.ts:42, which validates the cart, then calls..."If the user starts conceptually, redirect immediately: ask them to pick one concrete event and walk what happens, step by step, from there.
When the session closes, propose one follow-up:
/connect-to-what-you-know/quiz-me (optionally /flashcards to capture)/linked-notes/assessnpx claudepluginhub ic3b3rg/learn-skills --plugin learn-skillsPrompts developers to explain AI-generated code or plans via rubber duck questioning to verify comprehension and prevent rubber-stamping.
Retrieval-before-consultation quiz for active recall. User answers from memory before the agent verifies against code, docs, or a topic. Use to consolidate learning or check retention.
Verifies user comprehension of AI-generated results via categorization, gap detection (Expectation, Causality, Scope, Sequence), and Socratic questioning. Use when AI outputs remain ungrasped.