Fill-in notes — the Loom method

SkillDocs & knowledge

Turn a textbook chapter, lecture, or paper into beautiful "fill-in" study notes — written to be READ (clean statements, intuition) yet engineered to be FILLED (blanks, proof skeletons, "your turn" computations) so the reader learns by active recall. Notes are typeset with the Loom XeLaTeX class in this repo. Use when the user wants study notes from a source with gaps to fill in, worked examples restaged as exercises, or theorems stated but proofs left open. Triggers: "make fill-in notes", "turn this chapter into notes I can fill in", "guided / skeleton study notes", "study notes from this book/PDF", "读+填 笔记", "被动+主动 learning notes". NOT for a plain summary, a polished paper, or notes with every detail spelled out (that kills the active layer).

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the Fill-in notes — the Loom method skill

What this skill tells your AI

The instructions your AI receives, as published by polaris-aeterna/loom-notes in skill/SKILL.md and read by ahel’s review.

Produce study notes that a learner both reads (passive intake) and fills (active recall). The source is the answer key; the notes are the scaffold. Lead with a spine (one organizing idea) so a chapter becomes a single story, not a transcript.

Pipeline

  1. Read the source; find the spine. Skim the chapter/PDF, then name the ONE organizing idea and reorganize around it — a dictionary, an engine list, a single theorem the rest orbits. Do not transcribe section by section if a better shape exists. (Concentration → "8 engines"; Nonlinear Algebra → the algebra↔geometry dictionary.)
  2. Scaffold. Copy template/ (it brings loom.cls + a starter main.tex). Put the master spine on the front page (a cheat-table / map, partly blank), then \input one file per section under sections/.
  3. Draft each section in the Loom grammar (see the table below).
  4. Engineer the gaps (~70% read / 30% fill). Read reference/method.md. Blank the high-value thinking moves, never so much that reading breaks.
  5. Compile + verify. latexmk -xelatex main (or xelatex twice). Fix the known reference/pitfalls.md traps. Rasterize a page (ghostscript) and read it back to check the look before declaring done.

The Loom grammar — what goes where

beatdevicerole
one-line thesisstrand envthe big idea, read-only
when to use it\trigger{…}a cue line
sub-heading\block{…}quiet section lead
quick-referencetabularx w/ L columnthe cheat-sheet / dictionary
definitiondefinition env (madder knot)state it, blank a key clause with \fillin
resulttheorem/lemma/prop/cor (indigo knot)state in full (read)
proofproof + \TODO{step}skeleton; the learner fills the step
worked instanceexample (weld knot)setup given
do-it-yourselfyourturn env + \workspace[n]the active zone; the example, restaged
grok meter\warmth{0..5}self-assessment per section
margin recall\recall{question}a parked active-recall prompt
open thread\loose{…}an exercise to pull next time
cross-ref thread\warp{key} / \pick{key}declare and reuse a recurring object

Full command reference: reference/loom-commands.md.

Honest scoping

  • The notes restate results from the source. For personal study this is fine; before making notes that closely track a copyrighted book public, attribute clearly and consider an original example instead (see the repo README's note).
  • Keep the source's numbering in the knot title (e.g. Theorem 4.1 (… Thm. 2.16)); the bold auto-numbers are local to the notes.

See also

  • ../template/main.tex — blank starter (a live cheat-sheet of every device).
  • ../examples/nonlinear-algebra/ — full fill-in treatment (\fillin, yourturn).
  • ../examples/concentration-of-measure/ — proof-skeleton style (\TODO).
  • ../examples/demo/ — every visual feature on two pages.

Signals

GitHub stars
153
Forks
12
Last commit
Jun 2026
Advanced
Catalog kind
skill
Gateway key
fill-in-notes
Source
github.com/polaris-aeterna/loom-notes