agentleFS
calebnewtonusccalebnewtonusc/Chewbacca/skills/study-guide/SKILL.md

study-guide

Skill8 starsChanged 3 months agoMITView on GitHub

Build an interactive study guide, review sheet, practice quiz, walkthrough or flashcard set for a course topic, from lecture notes, slides, a transcript, a reading or a problem set. Produces one HTML file with quizzes, step-throughs and flashcards that records what was missed, so a later session picks up there. Use when asked to make or extend a guide, turn notes into practice, or review before an exam.

5 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 1 is distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 5 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    If guide progress shows missed topics, the new guide opens on those.

    line 33
  2. Distinctive:

    A wrong answer must be tempting: Every distractor should be something a person who half-learned this would actually pick: the right idea in …

    line 118

“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

Study guide

The study-system skill is the pedagogy: ask, do not tell; mark the gap specifically; bring the question back later. Its weakness is that a conversation ends and takes the retrieval practice with it. Next week starts from zero on material already failed once.

A guide is where that survives. One HTML file with quizzes, step-throughs and flashcards, plus a JSON sidecar holding per-topic results that guide progress reads back. Write markup only. The runtime is already written.

Before writing anything

guide list                       # extend an existing guide, do not start a rival
guide progress                   # what they already got wrong
coursework policy <course> ai    # what help this course actually allows

The policy check is not optional and it is not a formality. Courses differ hard: one bans AI outright, one requires disclosing every prompt, one allows ideation only. A practice quiz is usually fine where a drafted essay is not, but read the rule for the specific course before producing anything that reaches an assignment. If the course bans it, say so and offer to quiz them live instead, which is a thing you can do with no artifact at all.

If guide progress shows missed topics, the new guide opens on those. Do not rebuild the whole unit because one section was weak.

Making one

guide new "cache coherence" --course CSCI170   # writes guides/cache-coherence.html

Then edit the file: replace the starter blocks with real content and delete the ones you did not use. Then:

guide open cache-coherence

Always tell them to open it with guide open, never by double-clicking the file. Opened as a file:// URL the runtime can only reach localStorage, which nothing outside that one browser can read, so the next session cannot see what was missed and the entire reason the guide exists is gone. The page says which mode it is in, in the corner.

The markup

Quiz. One section.rg-quiz per concept. data-topic is a stable slug and progress is keyed on it, so renaming it orphans the history.

<section class="rg-quiz" data-topic="mesi-states">
  <h2>Check yourself: MESI</h2>

  <div class="rg-question" data-answer="c">
    <p>A line in the Shared state is written by this core. What state does it move to?</p>
    <ol class="rg-options">
      <li>Invalid</li><li>Exclusive</li><li>Modified</li><li>It stays Shared</li>
    </ol>
    <div class="rg-explain">A write needs ownership, so the other sharers are
    invalidated and the line becomes Modified.</div>
  </div>

  <div class="rg-question" data-answer="a,c">
    <p>Which states allow a silent read hit, with no bus traffic?</p>
    <ol class="rg-options"><li>Modified</li><li>Invalid</li><li>Shared</li></ol>
  </div>

  <div class="rg-question" data-answer="write-back|writeback">
    <p>What does a Modified line do on eviction?</p>
    <input class="rg-input" placeholder="one or two words">
  </div>

  <button class="rg-check">Check answers</button>
</section>

data-answer takes a letter (c), a 1-based index (3), a comma list for select-all (a,c), or |-separated alternatives for typed answers, matched case- and whitespace-insensitively.

Step-through. A procedure one step at a time, Prev/Next and arrow keys. Put the resulting state in each step, not only the action: the state is the thing they have to be able to reconstruct in an exam.

<div class="rg-steps">
  <div class="rg-step"><h3>1. Core 0 reads X</h3><p>Miss, goes to memory. Line enters <b>Exclusive</b> in C0.</p></div>
  <div class="rg-step"><h3>2. Core 1 reads X</h3><p>Snoop hit in C0, both now <b>Shared</b>.</p></div>
</div>

Flashcards and reveal:

<div class="rg-cards">
  <div class="rg-card"><div class="rg-front">Write-allocate</div>
    <div class="rg-back">On a write miss, fetch the line, then write into it.</div></div>
</div>

<details class="rg-reveal"><summary>Show the derivation</summary><p>...</p></details>

Writing good questions

This is the whole job. The markup is trivial and a guide full of bad questions is a worksheet with extra steps.

  • A wrong answer must be tempting. Every distractor should be something a person who half-learned this would actually pick: the right idea in the wrong direction, the adjacent term, the answer to the previous question. Three obviously-absurd options make a question that tests reading, not knowing.
  • Ask for the mechanism, not the label. "What does MESI stand for" is recognition. "A line is Shared and this core writes it, what happens to the other copies" is the thing the exam asks.
  • Use the typed-answer form for anything they must produce from nothing. Multiple choice always leaks the answer set, which is exactly the crutch the exam removes.
  • rg-explain explains the tempting wrong answer, not just the right one. Saying why C is correct teaches less than saying why B is the trap.
  • Cite sources in the footer, with the lecture date or slide range. A guide whose claims cannot be traced is worth nothing a week later when it disagrees with the professor.

Afterwards

guide progress                 # what is still wrong
guide progress <name> --json   # for another skill to read

Bring the missed items back in the next session before anything new. That is the whole loop, and skipping it makes the sidecar decoration.

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.