agentleFS
Agents4Academia-AIAgents4Academia-AI/UReKA/.claude/skills/tutor/SKILL.md

tutor

Skill9 starsChanged 5 months agoMITView on GitHub

Learn a course in scheduled hour-blocks — the tutor hands over a reading list of the real paper/blog/Wikipedia links, teaches the session's module, suggests ingesting what you read, then (if you opt in) drills adaptive flashcards or trusts the honor system, tracks mastery, and advances. Third stage of a course.

28 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use defuddle, obsidian-cli and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • HK-hubSkill

    defuddle

    Extract clean markdown content from web pages using Defuddle CLI, removing clutter and navigation to save tokens. Use instead of WebFetch when the user provides a URL to read or analyze, for online documentation, articles, blog posts, or any standard web page.

    Also in 4 projects that use this

  • jykimSkill

    obsidian-cli

    Interact with Obsidian using the official CLI to read, search, create, and manage notes, daily notes, tasks, properties, and more. Use when the user wants to query or manipulate vault content via the obsidian command.

    Also in 4 projects that use this

  • wangzhuofengSkill

    json-canvas

    Create and edit JSON Canvas files (.canvas) with nodes, edges, groups, and connections. Use when working with .canvas files, creating visual canvases, mind maps, flowcharts, or when the user mentions Canvas files in Obsidian.

    Also in 3 projects that use this

  • MicrosoftSkill

    excalidraw-diagram

    Create Excalidraw diagram JSON files that make visual arguments. Use when the user wants to visualize workflows, architectures, or concepts.

    Also in 3 projects that use this

  • wangzhuofengSkill

    obsidian-bases

    Create and edit Obsidian Bases (.base files) with views, filters, formulas, and summaries. Use when working with .base files, creating database-like views of notes, or when the user mentions Bases, table views, card views, filters, or formulas in Obsidian.

    Also in 3 projects that use this

  • GoogleSkill

    teach

    Teach the user a new skill or concept, within this workspace.

    Also in 4 projects that use this

  • +1 more from wangzhuofeng

The file

Show its rules and text

We split this skill into the rules it gives an agent

17 rules. 0 are common, 3 are distinctive, and 14 are very specific to this project.

  • 0

    Common

    Most of the 22 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 14

    Very specific

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

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

    If $ARGUMENTS is empty, glob course/*/goal.md, read each title, list the courses, and ask which to run.

    line 21
  2. Very specific:

    Read course/<slug>/goal.md, plan.md, schedule.md, progress.md, the modules/ docs, and the course/<slug>/library/ files (the fetched …

    line 28
  3. Very specific:

    If testing_mode is unset in goal.md frontmatter, ask how the learner wants their knowledge level updated:

    line 38
  4. Very specific:

    Write the chosen value (test | honor) to goal.md frontmatter.

    line 46
  5. Very specific:

    If tutor_prompt_mode is unset, also ask (governs /edit's re-test reminders):

    line 49
  6. Very specific:

    Write session-end | after-3-edits | never back to goal.md.

    line 56
  7. Very specific:

    Before teaching, give the learner the session's reading list of real sources to go read on their own (in Zotero / Notion / Obsidian) — the …

    line 73
  8. Very specific:

    After teaching, suggest in plain prose (a short list — no AskUserQuestion or pop-up box) that the learner ingest the session's material …

    line 96
  9. Very specific:

    review: every due item across all modules.

    line 135
  10. Very specific:

    drill: every item in the named module/topic, due or not.

    line 136
  11. Very specific:

    After scoring: show the relevant cited passage, give 1–2 sentences of feedback, and update the item's streak (confident → +1.

    line 168
  12. Distinctive:

    When an item is also backed by an ingested personal-base page, you may draw the follow-up/feedback from that page too — the knowledge level …

    line 171
  13. Very specific:

    If this session's module isn't mastered yet, recommend staying on it but let the user advance.

    line 177
  14. Very specific:

    Don't surface a downstream module's items until its prerequisite module is mastered (locked).

    line 178
  15. Distinctive:

    No flashcards. The learner has self-reported reading the session's sources and (Step 6) ingested them.

    line 182
  16. Distinctive:

    If a module crossed the mastery threshold this session, add an end-of-module milestone recap: module mastery %, deepest Bloom reached, a …

    line 210
  17. Very specific:

    When /edit finishes on a page it should respect the user's cadence:

    line 273

“Common” is measured: kept by at least 6 in 10 of the 22 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

Drive a course session by session for $ARGUMENTS, one scheduled hour-block at a time — the third stage of the /goal → /curriculum → /tutor pipeline. Each session runs the steps below: reading list → teach → suggest ingest → (test | honor) → recap → advance, tracking mastery in course/<slug>/progress.md.

Modes (parse from $ARGUMENTS):

  • <slug> — the next scheduled session (reading list + teach + ingest + test/honor + recap). Default.
  • <slug> hour <N> — run a specific scheduled hour-block instead of the next one.
  • <slug> review — spaced-repetition only: skip the reading list, teaching, and ingest prompt; drill all due items (test track only).
  • <slug> drill [module|topic] — on-demand flashcards for a module or sub-topic, ignoring due dates. Scores are still recorded.

If $ARGUMENTS is empty, glob course/*/goal.md, read each title, list the courses, and ask which to run.


Step 1 — Load the course

Read course/<slug>/goal.md, plan.md, schedule.md, progress.md, the modules/ docs, and the course/<slug>/library/ files (the fetched resources — their frontmatter holds the canonical URLs and credibility tiers you need for the reading list). Extract the target, tutor_prompt_mode, testing_mode, deadline, hours_per_*, the module map (the prerequisite chain), and the Item Progress + Module Mastery tables. If plan.md/progress.md are missing, offer to run /curriculum <slug> first.

Step 2 — Set preferences (first run only)

If testing_mode is unset in goal.md frontmatter, ask how the learner wants their knowledge level updated:

How should I gauge your progress each session?

  1. Test me — drill adaptive flashcards and score me (mastery is measured)
  2. Honor system — no flashcards; I'll ingest only material I've actually read and you trust that as progress

Write the chosen value (test | honor) to goal.md frontmatter. It can be overridden per session if the user asks.

If tutor_prompt_mode is unset, also ask (governs /edit's re-test reminders):

After you edit a page this course covers, when should I remind you to re-test?

  1. At session end — once per edit session, at the very end (recommended)
  2. After 3 edits — prompt after every 3 edits to covered pages this session
  3. Never — I'll run /tutor manually

Write session-end | after-3-edits | never back to goal.md. /edit reads it — see Tutor prompt trigger at the bottom.

Step 3 — Pick this session

(review and drill modes skip teaching and ingest — jump to Step 7.)

Get today's date (date +%Y-%m-%d) — used only for spaced-repetition intervals, not to gate the schedule (the learner studies whenever they like). In schedule.md, find the first un-ticked ## Week <W> · Hour <H> block; the module(s) it names are this session's target. For hour <N>, use that block. If every block is ticked, tell the user the schedule is complete and switch to review mode. From plan.md's Resources list (and the module's sources: frontmatter), gather this session's resources — the specific library files / links assigned to this hour-block's module.

Step 4 — Hand over the reading list

Before teaching, give the learner the session's reading list of real sources to go read on their own (in Zotero / Notion / Obsidian) — the actual links, not the library/.md files. For each resource, pull from its library/<file>.md frontmatter and present:

  • Title + canonical link — the url: (web) or source_links: arXiv/blog URL, clickable.
  • Kind · credibility tier (paper / Wikipedia / blog · canonical / reputable).
  • TLDR (one line) — why you recommend it and what the learner will get (e.g. "the original CFM objective + Gaussian/OT paths — read this for the derivation").
  • Quick-browse — a pointer to the library/<file>.md summary to skim if they don't want the full source now.

Frame it as go read and annotate these from the source; the summaries and lecture are your quick reference. Keep it tight — only this session's resources, ordered as the module uses them.

Step 5 — Teach

Present a tight, grounded synthesis of this session's scheduled module sections, read straight from course/<slug>/modules/<n>-<slug>.md — this is the "lecture", and the quick-browse companion to the reading list. Never invent: explain only what the cited module content covers, in the session's ~1-hour budget. Offer to go deeper on any part before moving on.

Step 6 — Suggest ingesting into the personal base

After teaching, suggest in plain prose (a short list — no AskUserQuestion or pop-up box) that the learner ingest the session's material into their personal base. For each resource:

  1. Suggest the matching helper command using the resource's real title, not the library slug — /zotero "attention is all you need" (PDF + annotations), /alphaxiv … (arXiv), or /web <url> (Wikipedia/blog). The learner reads and annotates the source themselves, then ingests those annotations.
  2. Dedup-check first: glob papers//concepts//sources//notes/ for a match. If one exists, suggest an update instead of a duplicate (re-ingest for fresh annotations, or re-run /collate//edit <page>) — naming the file it matches.

This is a suggestion, not a gate — continue to Step 7 without waiting, noting anything the learner says they ingested. (Invoke a helper only if they explicitly ask you to ingest for them.)

Step 7 — Test or honor (branch on testing_mode)

Track A — test: adaptive flashcards

Drill this session's items and update the knowledge level from both the curriculum material and what's now in the personal base (papers/sources/concepts/notes ingested in Step 6 or earlier).

7A.1 — Build the flashcard queue. Candidate items come from the Item Progress table. Order them with streak-graded spaced repetition — an item is due when days since Last tested ≥ its interval:

StreakRetest interval
01 day
13 days
27 days
3+14 days

Queue priority (first → last): wrong → partial → untested → confident & due. Confident items not yet due are skipped. Within a tier, oldest Last tested first.

  • This session: items from this session's module that are due, plus any resurfaced due items from mastered earlier modules.
  • review: every due item across all modules.
  • drill: every item in the named module/topic, due or not.

Adapt to mastery (Bloom escalation). Each item carries a Bloom level (remember → understand → apply → analyze → evaluate). Generate each flashcard live from the module content at the item's current level — but for items from a mastered module, pose the next level up (a remember definition becomes an apply/analyze prompt). Tell the user the queue size (~3 min/item); they can type "done" anytime to end early and still save scores.

7A.2 — Question loop. For each queued item, generate one flashcard from its module page, varying type across the session (don't repeat a type twice in a row):

TypeWhenExample
Definitionfirst-time / remember"Define a probability path $p_t$ precisely."
Mechanisma described process"Walk through how the CFM loss is computed."
Comparisontwo related items"How does the OT path differ from the diffusion path?"
Applicationpartial/revisit/apply+"When would you prefer rectified flow, and why?"
Mathspages with equations"Write $\mathcal{L}_\mathrm{CFM}$ and explain each term."

For maths, quote the LaTeX so the precision expected is clear. Accept: an answer → score it; "skip" → no score change, move on; "show" → display the cited passage, move on without scoring; "done" → end the session.

Score each answer:

ScoreCriterion
confidentCorrect and precise; no significant gaps
partialMostly right but missing a key detail or a minor error
wrongMissed the core idea or significantly incorrect

After scoring: show the relevant cited passage, give 1–2 sentences of feedback, and update the item's streak (confident → +1; partial/wrong → reset to 0) and its Bloom reached (raise it when the learner answers a higher-level prompt correctly). When an item is also backed by an ingested personal-base page, you may draw the follow-up/feedback from that page too — the knowledge level reflects curriculum and personal base.

7A.3 — Mastery gating (advisory). A module is mastered once ≥80% of its items are confident, each with streak ≥2 (the canonical rule lives in .claude/src/curriculum_utils.py). If this session's module isn't mastered yet, recommend staying on it but let the user advance. Don't surface a downstream module's items until its prerequisite module is mastered (locked).

Track B — honor: trust what was read

No flashcards. The learner has self-reported reading the session's sources and (Step 6) ingested them. Update the knowledge level from the new personal-base material alone, on the honor system — you trust that the learner only ingests material they've actually finished:

  • For each of this session's items now backed by an ingested page/source in papers//concepts//sources//notes/, mark its Item Progress Score: read (the honor marker — distinct from a tested confident), set Last tested to today, and leave Streak at 0.
  • Items with no ingested backing stay untested. Briefly note which items are still unbacked so the learner knows what's outstanding.
  • Honor read items count as untested in the goal audit (read, not tested) — not covered, which still requires a tested confident. Mention they can switch to test mode anytime to verify and raise true coverage.

Step 8 — Recap and next steps

Print a recap:

  • Covered: module + items learned/tested this session, and which sources were ingested.
  • Test track — Scores (confident / partial / wrong / skipped); shaky items (partial/wrong, each with a "re-read §X" pointer); spaced-rep forecast (which items are due next and when).
  • Honor track — items now read/ingested vs still outstanding; the read/ingested count alongside tested coverage.
  • Next session: the next un-ticked day from schedule.md and its module.
  • Suggest next steps: e.g. ingest a still-missing source, run /collate to fold new reading into a page, switch test/honor mode, or (test track) re-test shaky items.

If a module crossed the mastery threshold this session, add an end-of-module milestone recap: module mastery %, deepest Bloom reached, a "you can now…" tied to its learning objectives — then unlock the next module.

Step 9 — Write progress.md and sync

Update course/<slug>/progress.md:

  • Item Progress rows: new Score (confident/partial/wrong on the test track; read on the honor track), Bloom, Last tested (today), Streak. Carry over rows not touched this session unchanged.
  • Module Mastery rows: recompute Mastery %, Confident, Bloom reached, Status.
  • Frontmatter: recompute coverage_pct (confident items ÷ all items — honor read items do not count toward it) and course_mastery_pct (mean of module mastery %); set last_updated to today.

Then sync goal.md: overwrite the Knowledge Audit Status/Score/Last tested to match — covered for confident items, untested for honor read items, revisit for partial/wrong — and update its coverage_pct. Tick completed days in schedule.md and any finished modules in plan.md's weekly plan. Change nothing else.

Step 10 — Promote mastered material to the personal base

Course material lives in the per-course library course/<slug>/library/ until absorbed. At the end of the session (after the recap), identify library sources eligible for promotion. A course/<slug>/library/<file>.md is eligible when any holds:

  • Covered — backing items all confident (100% covered), even before streaks reach the mastered bar. Offer as soon as a module hits 100% confident; don't wait for the retest.
  • Mastered — its module is mastered (per Step 7A.3).
  • Ingested + annotated — the learner ingested it into the personal base (the honor-track promotion path).

For eligible sources, suggest promoting them into the personal sources/, noting which bar triggered it (e.g. "Module 1 is 100% covered — add these to your personal sources?"). On yes, cp course/<slug>/library/<file>.md sources/<file>.md (keep the library copy). Once in sources/ it enters the retrieval index, so a future /goal audit finds it without re-fetch and /collate//concept can synthesise it into a page. Never promote silently, and never below 100% covered (test) / ingested (honor). If the learner already ingested the real source in Step 6, that personal page is the durable copy — no library promotion needed.

Step 11 — Advance

Move the course forward: confirm this session's hour-block is ticked in schedule.md, and point the learner at the next hour-block's module (or the next module if this one is complete). Offer to continue now (/tutor <slug> for the next hour-block, or /tutor <slug> hour <N>), or to stop here. If the schedule is finished, suggest /tutor <slug> review to keep due items fresh.

Wrap up

Beyond the recap, promotion offers, and the advance pointer, ask:

  • (Test track) Re-test any wrong/partial items now?
  • Incorporate into your personal knowledge base (or update) a source you read today that isn't in the base yet?
  • Open a page that revealed gaps via /edit <page>?

Then respect tutor_prompt_mode for the next /edit (below). Do not reindex manually — the background poller handles it.


Tutor prompt trigger — spec for /edit integration

When /edit finishes on a page it should respect the user's cadence:

  1. Glob course/*/goal.md. For each, check whether the edited page's slug/title appears in its Knowledge Audit or the course's Module Map. If no match, do nothing.
  2. For a matching course, read tutor_prompt_mode:

    session-end — at the very end of the /edit wrap-up (once per session, even if several pages in the course are edited), append:

    Course <slug> covers this page — remember to run /tutor <slug> this session. Track with a flag file course/<slug>/.session_reminded (deleted on session start, created on first reminder).

    after-3-edits — read the integer in course/<slug>/.edit_count (0 if absent), increment, write back. When it reaches 3, prompt:

    You've edited 3 pages covered by course <slug> this session — run /tutor <slug>? Then reset .edit_count to 0.

    never or unset — no prompt.

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.