exam

# Exam Author an exam and its answer key, then stop. ## Intake Read [learning context](../learning-context.md). Use a learning brief supplied by `/learn` as the source; otherwise resolve the current conversation plus any argument and confirm one topic. In both cases, recall relevant prior knowledge at invocation and refresh the brief before asking exam setup questions. Ask these setup questions one at a time: 1. Format: multiple-choice, free-form, or mixed? 2. Length: how many questions? 3. Difficulty: introductory, intermediate, advanced, or a requested mix? Use recalled knowledge to avoid wasting questions on already-demonstrated material and to target prior gaps. Continue without personalization after reporting a recall failure. ## Write the exam Resolve the destination as: ```text ~/.agentic/artifacts/exams/<repo>/YYYY-MM-DD-<topic-slug>/ ``` Append `-2`, `-3`, and so on when needed. Resolve `<repo>` from the git root basename, then the current directory basename, then `no-repo`. Create: - `exam.md` containing the complete exam, - `answers.md` containing the complete answer key and concise grading criteria. Make every question answerable from the confirmed scope. Free-form questions should probe mechanisms, application, limits, or distinctions rather than rote feature lists. Multiple-choice distractors should represent plausible wrong models, not word games. Put answers only in `answers.md`; the file is visible and does not need an unlock step. Report both absolute paths and a concrete estimated completion time. ## Boundary This skill does not administer questions, wait for responses, grade answers, or write to agent memory. At completion, append the configured `exam` suggestions from [skill connections](../../../docs/skill-connections.md).