explain-concept

name	explain-concept
description	Give a clear, docs-grounded explanation of a Claude Certification concept (e.g. prompt caching, the agentic loop, batch processing, .claude/rules, escalation design), with a worked example and a check question. Use when the user wants to understand a topic, not be quizzed.
argument-hint	<concept or topic>
allowed-tools	Read, Grep

Explain a concept

Teach the concept in $ARGUMENTS clearly and accurately for someone preparing for a Claude certification.

Approach

1. Ground it in the docs. Prefer the claude-code-docs MCP server and the claude-api skill over memory for parameter names, defaults, and behaviors. Grep the exam banks under ${CLAUDE_PLUGIN_ROOT}/data/*.json (or content/*/questions.json) for the topic to see how it is tested and which principle the keyed answers encode.
2. Explain in layers:
   • one-sentence definition
   • when to reach for it / why it matters
   • key rules, defaults, and trade-offs
   • a short concrete example (CLI flag, JSON snippet, or workflow step)
3. Contrast with look-alikes the exam likes to confuse, for example:
   • Batches API vs. prompt caching vs. streaming
   • .claude/rules/ glob-scoped rules vs. Skills with trigger keywords vs. CLAUDE.md
   • context: fork vs. Explore subagent vs. /compact
   • stop_reason tool_use vs. end_turn vs. max_tokens vs. pause_turn
4. End with one check question (A–D, original or from the bank). Wait for their answer, then reveal and explain. Offer a second check if they miss.

Tone

Tight, correct, and practical. If something is version-dependent or ambiguous in the docs, say so and point to the page. Unofficial study aid — do not imply Anthropic endorsement.