← ClaudeAtlas

bld-quizlisted

qz: Post-sprint learning check. Quiz the user on what was just built, sized to the sprint; tiny changes get a walkthrough instead. Use for "quiz me", "test me on that", or when a sprint wraps.
mnguyenht/bld-package · ★ 1 · Web & Frontend · score 64
Install: claude install-skill mnguyenht/bld-package
# bld-quiz — turn the sprint just built into a learning checkpoint The user is learning by building, and Claude just wrote code they now need to actually understand. **Do not assume how much they already know.** Read it off the conversation: what they have asked, what they corrected, what they clearly already use without explanation. If the sprint touched something where the right depth is genuinely unclear, ask once, in one line, and then stay at that level. A quiz pitched at the wrong level teaches nobody: too low is patronising, too high just produces wrong answers and no learning. Turn this sprint's diff into a checkpoint sized to what changed. **Quiz only what Claude changed this sprint** — never pre-existing code, unrelated files, or another session's work. Keep it cheap: the diff is already in context from the sprint itself. Don't re-read the whole app, don't spawn agents, don't re-run `git log` archaeology. ## Step 1 — Size the sprint Judge by complexity and learning value, not raw line count: - meaningful lines changed — excluding lockfiles, generated files, and formatting-only churn, - relevant files touched, - conceptual difficulty, and how many concepts are genuinely new to the user, - whether architecture, data flow, state management, APIs, or the database moved. | Tier | Looks like | Checkpoint | |---|---|---| | Micro | one-line fix, copy change, CSS value, rename, config value | no quiz — walkthrough below | | Quick fix | one simple bug, ~2–10 meaningfu