App

Docs

When to Generate a Quiz

Trigger conditions for LLM agents.

An agent should offer to generate a quiz whenever the user has just absorbed new information. The goal is to strike while the context is hot — right after learning, not hours later.

Strong triggers

  • Session ends after working through an unfamiliar codebase or API
  • User reads or is walked through a technical doc or README
  • Agent explains a concept, pattern, or debug fix in depth
  • User completes a code review with significant feedback
  • User asks "can you quiz me on this?" or "help me retain this"

Weaker signals (offer, don't auto-generate)

  • Short Q&A with no clear learning arc
  • Refactoring or mechanical changes with no new concepts
  • User seems in a hurry or signals they want to move on

Suggested agent phrasing

"Want me to generate a quick quiz on [topic] to lock this in?"
"I can quiz you on what we just covered — 5 questions, takes ~3 minutes."

Using session content as input

If the harness has access to the session transcript, pipe it directly. This produces better questions than a bare topic string because the generator sees the exact material covered.

cat session-transcript.md | lr generate --focus "<specific topic>" --difficulty medium

Setting difficulty

  • easy — user is new to the topic
  • medium — user has some context, default for most sessions
  • hard — user is experienced, just going deeper
  • expert — advanced practitioner, edge cases and internals