agentleFS
baney75baney75/myelinate/skills/flashcards/SKILL.md

flashcards

Skill4 starsChanged 2 days agoMITView on GitHub

Use when someone says make flashcards, a study deck, or Anki or Quizlet cards. Quick live drills belong to tutor; cards support recall after understanding.

We split this skill into the rules it gives an agent

6 rules. 0 are common, 2 are distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 5 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Very specific:

    Stand down when it isn't wanted: If memory or the person says not to use Myelinate for this subject or kind of request, skip this skill's …

    line 12
  2. Very specific:

    Understand first, then card: If the learner can't explain it, teach it (tutor) before carding it.

    line 18
  3. Distinctive:

    Don't touch existing decks: add new ones; never overwrite or merge without asking.

    line 22
  4. Distinctive:

    Leave out: things they'll look up, things they'll never be asked, anything they can derive in seconds.

    line 59
  5. Very specific:

    Ask which tool they use if unknown.

    line 63
  6. Very specific:

    No tabs or line breaks inside a card.

    line 70

“Common” is measured: kept by at least 6 in 10 of the 5 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Flashcards

Cards exist for durable recall of what the learner already understands: fast, reliable access to facts, terms, relationships and steps so working memory is free for reasoning. A deck cannot create understanding, and a bad deck creates review debt and false confidence. Few, sharp cards beat many vague ones.

Paths like references/…, templates/… and examples/… are relative to this skill's folder, which carries its own copy of every file it names. Other Myelinate skills are siblings (../tutor/SKILL.md) and are installed as their own skills too. If a file is missing from your copy, fetch it from https://raw.githubusercontent.com/baney75/myelinate/main/ plus its path in the repository (references/…, skills/<name>/SKILL.md); if you can't fetch, work from this file and the router's core rules.

Stand down when it isn't wanted. If memory or the person says not to use Myelinate for this subject or kind of request, skip this skill's method and answer plainly; a slash command brings it back.

Evidence: retrieval practice and distributed practice are the two "high utility" study techniques (Dunlosky et al. 2013); repeated retrieval beats restudy at a week (Roediger & Karpicke 2006); successive relearning (3 correct recalls per item in the first session, then 1 correct recall in each of 3+ spaced sessions) raises college exam scores (Rawson & Dunlosky 2011; Rawson, Dunlosky & Sciartelli 2013). Card-writing rules below follow Matuschak ("How to write good prompts") and Wozniak ("20 rules of formulating knowledge").

0. Rules that override everything else

  1. Understand first, then card. If the learner can't explain it, teach it (tutor) before carding it.
  2. One retrieval per card. Atomic: one fact, one link, one step.
  3. Every back is checked against the course source or a reliable reference before delivery.
  4. Course wording and notation exactly as the course uses them; note where the textbook differs.
  5. Don't touch existing decks: add new ones; never overwrite or merge without asking.

1. What makes a good prompt

  • Focused: one detail. "Describe glycolysis" is twenty cards.
  • Precise: one acceptable answer. Add context to rule out reasonable alternatives: "In the proximal tubule, roughly what fraction of filtered Na⁺ is reabsorbed?" → about two-thirds (~65%).
  • Consistent: produces the same answer every review.
  • Tractable: answerable ~90% of the time once learned; break hard items down or add cue words.
  • Effortful: real retrieval. No yes/no. No front that gives the back away by pattern.
  • Short: back ideally one line. A paragraph back is several cards.
  • No orphans: each card serves an objective. Skip completionism.

2. Card types, mix them per concept

A concept deserves several angles; this is what turns cards into understanding scaffolds rather than trivia.

TypeFront → Back example
FactOn a topographic map, what does one contour line connect? → points of equal elevation
Mechanism / whyWhy does adding a resistor in series lower current at fixed voltage? → total resistance rises, so I = V/R falls
Cause → effectIn a basic supply-demand model, if supply increases and demand stays fixed, what happens to equilibrium price? → it falls
ContrastHow does a primary eyewitness account differ from a later historical synthesis? → direct testimony from the event versus an interpretation drawing on sources
ApplicationA history essay quotes a 1910 speech. Is that speech a primary source for the speaker's stated view? → yes
Procedure stepAfter choosing a common denominator for fraction addition, what comes next? → rewrite both fractions as equivalents with that denominator
Cloze (closed list/sequence)Water cycle: evaporation, {{c1::condensation}}, precipitation
Image occlusionMasked structure on a diagram → its name (one mask per card)
Definition ↔ termReverse only when both directions are tested (vocab, drug ↔ class)
Hook-assistedTarget fact on the front/back; the mnemonic (memory-hooks) as an optional hint line on the back, never the answer itself

Interference control: confusable pairs (primary/secondary sources, area/perimeter, mitosis/meiosis) get an explicit contrast card instead of two near-identical prompts.

Image occlusion: mask every instance of the label, image text, other labels that give it away by elimination, caption, alt text, filename. Use licensed or course-supplied images only (find-media).

3. How many and which

  • Per lesson: typically 6 to 15 cards. Per exam: weight cards by the study guide, not by what's easy to card.
  • Prioritize: high-yield facts the reasoning depends on, confusable distinctions, mechanisms' key "why"s, frequently tested values and steps.
  • Leave out: things they'll look up, things they'll never be asked, anything they can derive in seconds.

4. Export formats

Ask which tool they use if unknown. Default: Quizlet-ready block + readable table. Sets over ~20 cards: also deliver a file (.txt/.tsv/.md) and state the count.

Quizlet (Import → "Between term and definition: Tab", "Between cards: New line"):

Rate-limiting enzyme of glycolysis	PFK-1
Effect of ↓pH on O₂-Hb curve	Shifts right; more O₂ unloaded to tissue

No tabs or line breaks inside a card. If a back needs a line break, use a custom card separator (e.g., ;;) and say so.

Anki (File → Import; a plain-text file):

#separator:tab
#html:false
#tags column:3
Rate-limiting enzyme of glycolysis	PFK-1	BIOL126::Exam2::Metabolism

Cloze cards go in a separate file with #notetype:Cloze and text like Hemoglobin's Hill coefficient is about {{c1::2.7}}. Hierarchical tags Course::Exam::Topic.

Obsidian Spaced Repetition plugin: Question::Answer (one line), Question:::Answer (also reversed), multi-line cards with a ? line between front and back, ==highlight== cloze; separate every card with a blank line (a multi-line card ends at a blank line); tag the note #flashcards/<course>.

In-chat drill: show fronts one at a time; learner answers; reveal and rate (again / hard / good / easy); misses requeue at the end of the round; finish with the miss list as a mini deck.

Inside a lab (lab): built-in review with a stored schedule (e.g., 1 → 3 → 7 → 14 → 30 days on "good", 10 minutes on "again"), plus Quizlet and Anki export buttons.

5. Scheduling and use

  • Encourage an SRS (Anki/FSRS, Quizlet's spaced mode, Obsidian SR) over massed review.
  • First session: 3 correct recalls per card. Then relearn to 1 correct in each of 3+ later sessions, days apart; adjust spacing to the exam date (study-plan).
  • Say aloud or write the answer before flipping. Do not self-grade with "I knew that."
  • Pair cards with application practice; cards alone don't produce transfer.
  • Retire or rewrite cards that fail repeatedly (usually too big, ambiguous, or not understood). A leech is a teaching problem, not a discipline problem.

6. Quality check before delivering

  • Every back verified against source; numbers and units correct
  • Each card has exactly one answer; no yes/no; no giveaway fronts
  • Mix includes why/contrast/application, not only facts
  • No duplicates or near-duplicates; confusable pairs have contrast cards
  • Format imports cleanly (one tab per line for Quizlet; headers for Anki)
  • Count stated; tags/course labels applied

Anti-patterns

Paragraph backs · yes/no cards · cards before understanding · copying slide bullets verbatim as cards · 300-card decks for a 4-chapter exam · mnemonic as the answer · ambiguous prompts that accept several answers · reverse cards by default · overwriting the learner's existing decks.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.