---
name: markdownto-quiz
description: Author, repair, validate, and review portable quiz documents and their rendered results.
---
# Working with quiz
Use this skill when the user wants the quiz workflow. Read [SPEC.md](SPEC.md), then [shared conventions](../conventions.md). Published equivalents live at https://markdownto.ai/specs/quiz.md and https://markdownto.ai/specs/conventions.md. The spec and fixtures govern conformance; this editorial guidance is non-normative and adapts to the user's brief.
## Author and repair
Confirm the purpose from the request, inspect any supplied references, and preserve supplied content and constraints. Start from a relevant valid fixture, not a remembered grammar. Write the envelope and meaningful `##` sections, then sparse direction. Avoid invented IDs. Run `mdto validate file.md` after edits. Repair diagnostics at their reported location using the spec's rationale; preserve unrelated source. Prefer the spec-owned verbs for state changes and preview their diff when useful. If the CLI is unavailable, perform a clearly labeled manual check against the spec and do not claim tool validation.
## Editorial judgment
Define the learning objectives before drafting questions. Test reasoning with concrete situations, not recognition of your own wording. Write plausible distractors based on common mistakes, and keep alternatives parallel in length and grammar. Avoid trick questions, ambiguous negatives, and “all of the above.” Multiple-select questions should say to select every correct answer. Explanations should teach why the right answer works and why a tempting alternative fails. Verify every key independently and align points with learning importance. Test unanswered, partially answered and completed attempts, key order invariance for multiselect, edited answers after grading, and saved/reopened results. Be clear that this is self-study with source-visible answers, not tamper-resistant assessment.
## Review and delivery
Use `mdto render file.md` and the spec's build or review verbs. Inspect the actual resulting interface or artifact, including mobile/narrow views, keyboard controls and content extremes. Verify state-changing UI actions against the same CLI operations. Deliver the Markdown and usable output, explain any real limitation, and never equate a valid file with a successful artifact review. If artifact inspection is unavailable, say which checks remain undone.
Give this guide to your agent, or read it as an author. It covers workflow and editorial judgment; the specification defines conformance.
Use this skill when the user wants the quiz workflow. Read SPEC.md, then shared conventions. Published equivalents live at https://markdownto.ai/specs/quiz.md and https://markdownto.ai/specs/conventions.md. The spec and fixtures govern conformance; this editorial guidance is non-normative and adapts to the user's brief.
Author and repair
Confirm the purpose from the request, inspect any supplied references, and preserve supplied content and constraints. Start from a relevant valid fixture, not a remembered grammar. Write the envelope and meaningful ## sections, then sparse direction. Avoid invented IDs. Run mdto validate file.md after edits. Repair diagnostics at their reported location using the spec's rationale; preserve unrelated source. Prefer the spec-owned verbs for state changes and preview their diff when useful. If the CLI is unavailable, perform a clearly labeled manual check against the spec and do not claim tool validation.
Editorial judgment
Define the learning objectives before drafting questions. Test reasoning with concrete situations, not recognition of your own wording. Write plausible distractors based on common mistakes, and keep alternatives parallel in length and grammar. Avoid trick questions, ambiguous negatives, and “all of the above.” Multiple-select questions should say to select every correct answer. Explanations should teach why the right answer works and why a tempting alternative fails. Verify every key independently and align points with learning importance. Test unanswered, partially answered and completed attempts, key order invariance for multiselect, edited answers after grading, and saved/reopened results. Be clear that this is self-study with source-visible answers, not tamper-resistant assessment.
Review and delivery
Use mdto render file.md and the spec's build or review verbs. Inspect the actual resulting interface or artifact, including mobile/narrow views, keyboard controls and content extremes. Verify state-changing UI actions against the same CLI operations. Deliver the Markdown and usable output, explain any real limitation, and never equate a valid file with a successful artifact review. If artifact inspection is unavailable, say which checks remain undone.