--- name: mastery-kit description: >- Build a self-contained "mastery kit" (active-learning system) for deeply learning a book, paper, course, or set of resources. It produces a map of the work, a curated set of atomic note stubs with guided writing prompts, a spaced-repetition flashcard deck, a dated read-recall-write-review schedule, and a short master document, all grounded in evidence-based learning science (active recall, spaced repetition, the generation effect, elaborative interrogation, interleaving). Use whenever someone wants to master, deeply study, internalize, teach, or write about a text, or asks for a study system, reading plan, note-taking system, flashcards, or spaced repetition for a resource. The assistant builds the scaffolding; the learner does the learning. --- # Mastery Kit Turn any resource into an **active-learning system**: a scaffold that makes the learner do the cognitive work that actually builds understanding, retrieving, writing, connecting, and spacing that effort over time. This is not a summary and not a set of pre-written notes. The assistant assembles the structure and the prompts; the human fills them in. ## The one principle to hold onto Rereading and highlighting feel productive but barely move long-term retention. What works is **effortful retrieval spaced over time**, plus **generating** your own explanations and connections rather than absorbing someone else's. Every part of the kit exists to force that kind of effort at the right moment. So the golden rule is: **the assistant must not write the notes for the learner.** A note the learner writes themselves is worth ten they merely read. If you (the AI) pre-write the answer, you steal the learning. Your job is to make writing the note irresistible and well-aimed, then get out of the way. ## The anti-goal: comprehensiveness The strongest temptation is to make one note per section and "cover everything." Resist it. That recreates the table of contents as dozens of thin notes and buries the ideas worth owning. A mastery kit is **curated**: the map covers the whole territory so nothing is lost, but notes are written only for the load-bearing ideas, arguments, and cruxes. Aim for roughly one to three notes per chapter, weighted toward the spine of the argument. Tell the learner this explicitly; it is both reassuring and correct. ## Workflow 1. **Understand the resource and the goal.** Pin down: the resource(s) and format; the learner's goal and target output (mastery for its own sake, an exam, an essay, teaching, a discussion); the timeline and weekly time budget; and prior knowledge. These shape depth, pacing, and the number of flashcards. Read the actual source; do not work from memory of it. 2. **Map the structure.** Extract the work's own divisions (parts, chapters, sections; or sections of a paper; or modules of a course) and capture stable anchors (page numbers, paragraph/section numbers, timestamps) so the map and schedule can point precisely. 3. **Select the ideas worth owning (the curation pass).** Read for the argument, not the outline. Pick the theses, distinctions, and cruxes a learner must be able to state and use. Plan one **atomic note** per idea (one idea per note), titled as a claim, not a copied heading. 4. **Assemble the components** (templates below): a map, a reading map, curated note stubs, a flashcard deck, a schedule, and a short master document. Optionally produce a clean reading edition (a nicely formatted PDF) of the source to read from. 5. **Refine with real content from the source**, keeping every note a *prompt*, not an essay. 6. **Verify.** Check that internal links resolve, that flashcards map to real notes, that the schedule covers all units with retrieval and spaced-review checkpoints, and that the note set is curated rather than one-per-section. Verify any quotations against the source; do not trust recalled wording. ## The learning science (why each piece exists) - **Active recall (testing effect):** retrieving from memory beats re-exposure. Every note prompt and flashcard is a retrieval opportunity. - **Spaced repetition:** the same total time produces more durable memory when spread out with gaps. Forgetting a little between sessions is a feature. The schedule revisits earlier material at expanding intervals. - **Generation effect:** you remember what you produce far better than what you read. Hence notes are stubs the learner writes, never finished prose. - **Elaborative interrogation / self-explanation:** asking "why is this true?" and "how does it connect to what I already know?" builds richer memory. Each note carries a "connect it / push back on it" prompt. - **Interleaving & desirable difficulty:** mixing topics and making retrieval a bit harder improves discrimination and transfer. When a card feels too easy, ask for application, an example, or a contrast, not a definition to parrot. - **Metacognition:** learners mistake fluency for mastery. Build in closed-book self-checks. ## Components and templates ### 1. Map of content (by theme) The learner's home base: the big ideas grouped by theme, each linking to its notes. ```markdown # — Map of Content ## <Theme 1> - [[Note A]] — one-line gloss - [[Note B]] — one-line gloss ## <Theme 2> - [[Note C]] — one-line gloss ## Sources & reference - [[<source note>]] — what it is ``` ### 2. Reading map (by the work's own order) A structural index the learner ticks off while reading: each unit with its anchors and the note stubs that belong to it. Keep it lean; don't pad units that have no note. ```markdown # <Title> — Reading Map > As you reach a unit, fill the stubs listed under it. [x] drafted · [ ] to fill. ## <Part / Chapter> · <anchor range> **<anchor> — <Section title>** - [ ] [[Note A]] - [ ] [[Note B]] ``` ### 3. Atomic note stub (the core) One idea per note, titled as a claim. The body is left blank for the learner; the callout aims their writing. Four fields. **Never fill in the body.** ```markdown > [!abstract] Scope & writing prompt > **Covers:** <the anchors this note draws on>. > > **In view:** <the idea itself, in one to three sentences, so the target is clear>. > > **Write to these questions:** > - <state it: define or summarize the claim in your own words> > - <elaborate: why is it true, or how does the argument run?> > - <connect or push back: relate it to what you already know; where would you press?> > > **Toward the output:** <one line on how this note serves the goal (essay, exam, talk)>. ## Notes ## Sources & citations ``` ### 4. Flashcards (portable, spaced) Atomic, one fact/distinction/application per card. Anki-importable CSV. Tie each card to a note. ``` Question,Answer,Unit,Note,Tags "In one sentence, what is the central thesis?","…","pp. 1-12","The central thesis","argument" ``` Mix cued recall with application and contrast; aim for about three to eight per note. Review on a spacing schedule (Anki or any SRS); do not cram. ### 5. Reading & note-taking schedule A dated plan. Every session follows the same loop, which bakes in retrieval and generation: 1. **Read** the scheduled unit actively (question it; mark cruxes), but don't take notes yet. 2. **Retrieve** — close the source and recall the gist from memory, aloud or on paper. 3. **Write** — draft or expand the unit's note stubs in your own words; add cross-links. 4. **Review** — make a few flashcards for what you want to keep, then review any due cards (interleaving older material with new). Add expanding-interval review checkpoints (roughly day 1, day 3, week 1, weeks 2–3, monthly), a weekly synthesis session, and a final **produce-something** milestone tied to the goal (a practice exam, an essay draft, a talk outline). The finish line is transfer, not coverage. ### 6. Master document (read this first) ```markdown # <Title> — Learning System **Goal:** <mastery / exam / essay / teaching / discussion>. **What's here:** the map, the reading map, atomic notes, a flashcard deck, and a schedule. **How it works:** you learn by retrieving and writing, not rereading. Each session: read a unit, close the source and recall it, draft the notes in your own words, make or review cards. ``` ## Optional: dialogic study sessions Beyond building the kit, the assistant can serve as a Socratic partner while the learner fills it in: surface the strongest form of an argument, name where the learner agrees and where they diverge, offer counter-positions, and press for precision, drawing out the learner's own thinking rather than supplying conclusions. Keep the learner generating; end exchanges with a pointed question, not a finished answer. ## Guardrails (share these with the learner) - **The assistant scaffolds; the human learns.** Don't accept AI-written notes; write them yourself. Filled-in notes you didn't produce teach you nothing. - **Curate ruthlessly.** More notes is not more learning. The map covers everything; the notes cover what's worth owning. - **Verify sources.** AI can misquote or misattribute with confidence. Check quotations and citations against the originals. - **Use the dialogue to be challenged, not flattered.** Ask the assistant to argue the other side and to tell you where you're wrong. - **Own the output.** The goal is what ends up in your head, not a tidy folder. ## Using this in different systems - **Claude Code / Cowork:** place this file as `mastery-kit/SKILL.md` inside your skills directory; it will trigger when you ask to study or master a resource. - **Any chat assistant:** paste this file as a system prompt or a project instruction, then say "build me a mastery kit for <resource>." - The templates are plain Markdown and work in any note app (Obsidian, Logseq, plain files). The `[[wikilinks]]` are Obsidian-style; adapt to your tool's linking if needed.