chapter-digest
Skill1 starsChanged 51 days agoMITView on GitHub
Concise but complete summaries of sections, chapters, or page ranges from textbooks and reference books (for example CLRS, math and engineering texts), with nothing important dropped. Use this skill whenever the user shares a section, chapter, PDF, photo, or pasted text from a book and asks to summarize, condense, digest, "give me the key points", or says they have no time to read it all, even if they do not say "digest".
In this repository4 other files it gives its agents.
- exam-prepSkill
- math-solverSkill
- programming-studySkill
- socratic-tutorSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
19 rules. About 0 are common, 18 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈18
Distinctive · estimate
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
- Distinctive: line 16
If they only name a chapter without providing it, say that you have not been given the text.
- Distinctive: line 17
Never present anything as coming from the book unless it is in the text provided.
- Distinctive: line 21
Read through once and list, in order:
- Distinctive: line 23
Every heading and subheading.
- Distinctive: line 24
Every definition, theorem, lemma, property, rule, and law.
- Distinctive: line 25
Every formula and equation that is used later.
- Distinctive: line 26
Every algorithm, procedure, or method.
- Distinctive: line 27
Every worked example and its type.
- Distinctive: line 28
Every warning, note, or "common mistake" the author points out.
- Distinctive: line 41
If the user does not say which, use standard and mention the option to shrink or expand.
- Very specific: line 45
Follow references/digest-template.md for the exact layout.
- Distinctive: line 59
Every inventory item appears in the digest.
- Distinctive: line 60
Every formula, constant, bound, and condition matches the source exactly.
- Distinctive: line 61
Every algorithm keeps its steps and its complexity or convergence claim.
- Distinctive: line 63
If the book seems to contain an error or a confusing statement, do not quietly change it.
- Distinctive: line 68
Keep the book's notation so the reader's notes and the book line up.
- Distinctive: line 69
Keep the author's order of ideas, because it is easier to review against the book.
- Distinctive: line 76
Do not over-explain things the user has said they already know.
- Distinctive: line 82
After delivering, offer follow-ups that build on the digest: a quiz from the self-check questions, flashcards, or feeding it into an exam …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Chapter Digest
Purpose
Reference books are long and readers have little time. They need a summary they can trust to cover everything that matters, in a fraction of the length. Concise means cutting padding and repetition, never cutting a definition, theorem, formula, algorithm, condition, or exam-worthy point. When in doubt about an item, keep it and make it shorter.
Step 1: Get the actual text
- Work from the text the user provides. If a file is uploaded and its content is not already visible, read it fully with the file-reading or pdf-reading skill (if available) before writing anything.
- Note the page range and the section headings you can see. Check for unreadable pages, bad scans, or missing figures, and tell the user which parts you could not read instead of guessing.
- If they only name a chapter without providing it, say that you have not been given the text. Offer either to work from an upload, or a general overview from your own knowledge, clearly labeled as such, since it may not match their edition.
- Never present anything as coming from the book unless it is in the text provided.
Step 2: Inventory before summarizing
Read through once and list, in order:
- Every heading and subheading.
- Every definition, theorem, lemma, property, rule, and law.
- Every formula and equation that is used later.
- Every algorithm, procedure, or method.
- Every worked example and its type.
- Every warning, note, or "common mistake" the author points out.
- Summary or key-terms boxes and chapter-end exercises, if present.
This inventory is the coverage checklist. Nothing on it may silently disappear from the digest.
For long sections (more than about 15 to 20 pages), process in chunks of that size: inventory and digest each chunk, then merge into one document with a single outline. Do not stretch one pass over a whole long chapter, because details get lost.
Step 3: Choose the depth
- 1-page: the core ideas, the key formulas and algorithms, and the top mistakes. For a quick pre-class or pre-exam glance.
- Standard (default): about 10 to 15 percent of the original length. Complete outline, every inventory item, shortened examples.
- Detailed: about 25 to 30 percent. Adds proof ideas, more examples, and derivation sketches.
If the user does not say which, use standard and mention the option to shrink or expand.
Step 4: Write the digest
Follow references/digest-template.md for the exact layout. In short:
- Big picture (about 5 lines).
- Outline following the book's own headings, in order.
- A "nothing dropped" table of definitions, theorems, formulas, and algorithms, each with a one-line "use it when".
- Worked examples reduced to method (steps, key trick, result).
- Traps and must-remember box.
- Self-check questions with answers.
- Coverage check.
Adapt the content to the subject using references/formats-by-subject.md (algorithms, math, numerical methods, digital design, programming).
Step 5: Verify before delivering
- Every inventory item appears in the digest. Fix gaps, or list them in the coverage check with the reason.
- Every formula, constant, bound, and condition matches the source exactly. Recheck signs, exponents, indices, and inequality directions.
- Every algorithm keeps its steps and its complexity or convergence claim.
- Nothing has been added from outside the text. If extra help is useful, put it in a separate box labeled "Extra (not from the text)".
- If the book seems to contain an error or a confusing statement, do not quietly change it. Flag it to the user.
Fidelity and wording
- Paraphrase in your own words and keep quotations to short formal definitions or statements. State theorems and definitions accurately, with all their conditions.
- Keep the book's notation so the reader's notes and the book line up. Note when you introduce a shorthand.
- Keep the author's order of ideas, because it is easier to review against the book.
Style
- Friendly and direct in the chat message around the digest; match the user's tone.
- The digest itself should be clean and scannable: short lines, tables, and numbered steps instead of long paragraphs.
- Plain-text math in chat (x^2, sqrt(x), O(n log n)) so nothing breaks if LaTeX fails to render. LaTeX or proper typesetting is fine inside a PDF.
- Do not over-explain things the user has said they already know.
Delivering
- A 1-page digest can go in the chat.
- A standard or detailed digest works best as a document to keep (PDF, or Word if requested), made with the matching document skill if available.
- After delivering, offer follow-ups that build on the digest: a quiz from the self-check questions, flashcards, or feeding it into an exam study plan or cram pack.
Example
User uploads: a 16-page section of CLRS on heapsort and says "summarize this, I have no time".
You: Got it, 16 pages 💪 I'll do a standard digest (about 2 pages) and keep every definition, procedure, and complexity result.
Digest includes: the heap property and array index formulas, MAX-HEAPIFY, BUILD-MAX-HEAP, HEAPSORT each in short pseudocode with its running time, the reason building a heap is O(n) and not O(n log n) in two lines, a comparison table with other sorts, common mistakes (off-by-one indexing), five self-check questions, and a coverage list of every heading in those pages.
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
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.

