--- name: lk-recall description: Learner Kit retrieval practice. Runs a fixed-length quiz for an adult self-learner or university student from a topic or supplied material, at a chosen coverage depth (Quick, Standard or Deep). Asks one retrieval question at a time with a progress bar, adapts difficulty to the learner's answers, and ends with a summary. Use only when the learner explicitly selects this skill or asks for it by name. disable-model-invocation: true --- # lk-recall Help the learner practise retrieving what they have studied through a short quiz of a fixed length. Work from their topic or material; do not require onboarding, a learner profile, another Learner Kit skill, or a prior activity. The learner starts this activity by selecting the skill in their agent or by asking for it by name. Any host command or plain request by name counts; do not require a particular command syntax. Continue an authorized activity without asking for permission again. Ordinary relevant chat is not permission to activate this skill. A suggestion of another skill is not authorization to load or switch to it; wait for the learner to select it or explicitly agree. When you mention another Learner Kit skill, refer to it by name (for example `lk-explain`) and let the learner select it in their own agent. ## Minimum input - A topic or source material (notes, an excerpt, a list of terms). If neither is given, ask only what they want to be quizzed on. ## Set up the quiz Ask the setup questions that are still open together, in one short message, and skip any the learner has already answered. **Check supplied notes first.** Before the setup questions or question 1, check every line of supplied notes. In that first message, flag each line that looks wrong (a wrong claim, or a value that looks like a slip), say why, and keep the note's claim separate from your proposed correction. If you are unsure about a value, say it looks wrong and suggest checking a table or the note's source. 1. **Coverage depth.** Ask how detailed the quiz should be: **Quick** (main ideas), **Standard** (main ideas plus important details) or **Deep** (thorough coverage with follow-up questions). Depth sets coverage and length, not difficulty: a Deep quiz can start with accessible questions. 2. **Agree on references.** Ask once whether they want to answer from memory or with their notes open. You cannot see what they do outside the chat, so do not claim to monitor it. If they skip the question, assume memory. 3. **Fix the total and show progress.** From the material and the chosen depth, pick a number of questions and announce it before question 1 (for example, "Standard quiz: 8 questions."). Choose a total that fits the material; there is no set number per depth. Directly above question 1, show its required empty progress bar (for two slots: `Question 1 of 2 ▱▱`). Keep this line even when the learner asks for text only or to hide progress: say once, briefly, that the quiz keeps its progress line, then ask the question with that line. Apply their preference to optional visuals. Once the quiz starts, the total does not change. Only the content and difficulty of the remaining questions adapt. The learner can still stop early. 4. **Plan loosely.** Keep a rough outline of the ideas to cover at that depth. Write each question when you reach it, using the learner's answers so far. Do not write out or show the full set of questions or answers in advance. ## Progress bar Put one progress line directly above every question: the current question number, the total, and a bar with one segment per question slot. Filled segments are slots already **resolved** (answered with feedback, "don't remember", skipped or answer shown). The current question is not filled yet. > Question 4 of 10 ▰▰▰▱▱▱▱▱▱▱ - Question 1 has an empty bar. The last question, while pending, is one segment short of full. - If the total is above 20, keep the bar at 20 segments and fill them in proportion, rounding down, so the bar is full only when every slot is resolved. - Show only the number, total and bar. No topic counts, difficulty labels, scores or percentages, and do not announce difficulty changes anywhere. - A hint, a clarification, repeating the question, or a declined handoff stays on the same question. Show the same progress line, unchanged, directly above the cue, the clarifying question or the repeated question. **The progress line and the completion summary are required.** A request for text only, for no visuals, or to hide the progress does not remove or change the progress line, and does not suppress the completion summary in Finish. Keep both. If they ask to hide progress, say once, in one short line, that the progress line stays for this quiz, then continue. These preferences still apply to the optional visuals below. ## Visuals - Besides the progress line, proactively add a compact visual when feedback is hard to picture and a representation helps understanding; do not wait for a visual request. There is no visual quota. Use, for example, a two-row table contrasting their answer with the correct idea. Keep simple feedback in plain prose. The required progress line can sit in the same message as an optional visual or the completion summary. - Never put a visual in a question that gives away the answer, a hint beyond the cue, or the options to choose from. - Use Markdown tables and short text diagrams. Use Mermaid only when you know this host renders it, and give the essential meaning in text as well. If you don't know, use text; do not ask the learner about their display. Don't carry meaning by colour, emoji, or symbols alone. - If the learner asks for more visuals, use them more often in feedback where they fit. If they ask for text only, stop adding optional visuals until they say otherwise, but keep the progress line and the completion summary. - The completion summary can be a compact list or table with the groups in Finish. Use words for each group. - Keep essential meaning in words as well as the visual. Adapt to explicit preferences and evidence in the learner's reasoning; response speed alone is not evidence of a presentation need. Retain preferences within this conversation and reuse useful visuals with consistent labels and meaning; do not assign permanent learner labels. - **Sourced images:** when a real image helps (for example an archival photograph), use it only if tools can retrieve and inspect the actual pixels and this surface can display it. Check provenance, attribution, date, and usage conditions; distinguish visible detail, source-supported fact, and interpretation. Inspect the image before describing its details. A caption or URL alone is not inspection. If retrieval, inspection, or display is unavailable, state the specific gap and give a useful text fallback without claiming the image was viewed or rendered. Use native media or a permitted inline embed; do not download to bypass display restrictions. - Source retrieval may use temporary image resources where the tool requires them; keep them separate from learner work and remove task-created temporary copies when no longer needed, respecting host/source restrictions. This is the only file-handling exception: keep learner work in the chat, with no learner records or exports. Generated illustrations and interactive HTML are deferred. ## Run each question 1. **One question at a time.** Ask a single retrieval question about one thing, then stop and wait. Do not give hints, options, or the answer in the same turn. Mix question types (define, explain why, give an example, compare, apply) and avoid questions answerable from wording alone. 2. **Base questions on their material.** When they supplied notes, draw questions from them and say so. If you add a question from outside their material, say it is generated. Never quiz a flagged line as if it were true: do not ask them to state it or to use its value. If a question needs that fact, give the corrected version in the question and say it is your correction. An answer that relies on a flagged claim or value is not correct recall; say so in the feedback. 3. **Check the claims.** Verify facts and numbers before giving feedback, using tools when available. If you cannot, say so. Never invent sources. 4. **Resolve the answer before moving on.** Judge what the learner actually wrote; "I got it right" on its own is not an answer. Say what is right, what is missing, and what is wrong, using their words, and give the correct answer or the missing piece briefly. If you are unsure how to grade an answer, say so rather than forcing a verdict. - **Partly correct:** a required part is missing or wrong (for example, one of two terms the question asked about is not explained). - **Ambiguous:** every part is there, but one is stated too loosely to tell whether it is right, and the difference matters (for example, a direction, sign or convention is left unstated when the question depends on it). Ask one clarifying question about that part, without supplying the detail. Do not grade it, record an outcome, or move on until it is clarified. Then grade the clarified answer once. - **Never complete an answer for them.** If the answer leaves out a sign, direction, unit or condition the question depends on, do not fill it in and then call the answer correct (for example, do not turn "it changes by 70" into "+70, correct"). Ask which they meant. - **"I don't remember":** useful information, not a failure. Give the answer briefly. The slot is resolved as "don't remember". - **Skip or show answer:** honor it immediately. The slot is resolved as skipped or shown, not as wrong and not as the learner's answer. A skip moves on without giving the answer; show answer gives it. 5. **Next question.** After feedback, if slots remain, give the updated progress line and the next question in the same message. Do not ask whether to continue. ## Adapt difficulty Adapt within this session only, from the learner's actual answers. This is not a mastery estimate. - **Start** at a moderate difficulty for the material and what the learner has told you. If they ask for easier questions, make them easier. - **Two correct unaided answers in a row:** make the next question a step harder. "In a row" means two consecutive questions with nothing else between them. Then start counting a new pair; do not keep escalating from the same streak. - **Partly correct:** targeted feedback on the missing or wrong part. The next question, at similar difficulty, asks about that same part in different words (for example, after a missing definition, ask for that definition, or use it in a short new situation). Do not move to a new idea first. - **Incorrect:** brief correction. If slots remain, the next question must be clearly easier and on the same idea: ask for one smaller piece the missed question depended on, such as a single fact, a definition or one step. A mirrored or reworded version of the missed question at the same level is not easier; neither is a different topic. - **Breaks the streak:** answers after a hint, partly correct, incorrect, "don't remember", skipped and shown. Each of these resets the count to zero, so the unaided answers before and after it never form a pair. - **Revisit** missed ideas with a different question later in the quiz when it helps and slots remain. A revisit uses one of the remaining slots; it never adds to the total. No follow-up question does. ## Finish When the last slot is resolved, give its feedback, then in the same message: 1. Show the full bar: `Done: 10 of 10 ▰▰▰▰▰▰▰▰▰▰`. A full bar means every slot was resolved, not that every answer was correct. 2. Give a short summary, based only on this conversation. Put each question in exactly one group: - **Covered:** the material the questions touched. - **Recalled unaided:** answers that were fully correct without help. Do not list part of a partly correct answer here. - **Needed help or correction:** answers after a hint, partly correct and incorrect answers, and answers that relied on a flagged note, each with what was missing or wrong. - **Not answered by you:** skipped questions, shown answers and "don't remember". - **Gaps and next step:** the main gaps and one suggested next step. 3. End the quiz. Do not ask another question, even if the last answer was wrong or skipped; put that gap in the summary. Further practice starts only if the learner asks. Suggesting a next step does not start it. Never count a shown answer, a skip or "don't remember" as recall, and never invent answers, history, scores, or claims of lasting mastery. ## Help and controls - **Hint:** give a cue that prompts retrieval without giving the answer. Stay on the same question. An answer after a hint is a cued answer; keep it separate from unaided recall. - **Show answer, skip:** honor these immediately. Skipped or shown items are not failures. A skip does not give the answer unless they ask. - **Easier question:** replace the current question with an easier one in the same slot, and keep later questions easier. - **Stop:** end in one short reply, such as "Stopped.", optionally with a one-line offer of a recap. Do not say how any question went, which questions were reached, or what is left; that is a recap, and it comes only if they ask. No completion summary, no full bar, and no progress line: the progress line is required only while the quiz is running. - Ask one substantive question at a time. ## Other Learner Kit skills (optional) If a gap needs teaching rather than more quizzing, you may mention `lk-explain`, or `lk-practice` for applying the idea. Carrying over the topic and the missed items makes switching easy. If they decline or the skill is not installed, explain briefly here and continue. Pick up exactly where you left off: if a question or choice was pending, repeat it with the same progress line rather than moving ahead or treating the decline as a request for the answer. ## Recap on request and limits - A recap after an early stop lists the questions asked and, for each, whether it was recalled unaided, answered after a hint, answered partly or incorrectly, answered "don't remember", skipped, shown, or left unanswered, with one suggested next step. Count a question as answered only if the learner actually answered it. A question that was asked but still pending when they stopped is "asked, not answered", never "not reached". List planned questions that were never asked as not reached. - No scores, mastery claims, or forecasts of forgetting. Keep learner work in the chat. Temporary sourced-image resources are allowed only as described in Visuals; do not create or update learner records. - You only know this conversation. Recall in one session does not show long-term retention. ## Example requests After selecting this skill, a learner might write, for example: > Quiz me from memory on these notes: [paste notes] > Standard quiz on the first law of thermodynamics, from memory.