- Home
- Skills
- Research & Analysis
- literature-notes: reading notes for academic research
More screenshots
Works with the AI Agents you already use
literature-notes: reading notes for academic research
8 agent skills that turn academic PDFs into a structured, queryable wiki of reading notes anchored to pages, citations and your research scope.
$19
literature-notes: reading notes for academic research
Example session with this skill installed
/literature-notes-annotate ~/papers/vaswani-2023-attention-is-all-you-need.pdf --pages 5 --focus "multi-head attention"
- Read your context and instructions
- Compiled the literature-notes
Summary
Read page 5 of Vaswani et al. (2023) and wrote 1 reading note, anchored to the page and cited in APA 7, the norm chosen at setup. The source record already existed, so only the note was created.
Created
| File | Kind | Page | Tags |
|---|---|---|---|
notes/vaswani-2023--p5--multi-head-attention.md | summary | 5 | attention, transformer |
The note
---
type: Reading Note
source_id: '[[vaswani-2023]]'
page: 5
note_kind: summary
tags: [attention, transformer]
citations:
apa7:
text: (Vaswani et al., 2023, p. 5)
---
> "Multi-head attention allows the model to jointly attend to information
> from different representation subspaces at different positions."
## Configuration used
- h = 8 parallel attention heads.
- dk = dv = dmodel / h = 64 per head, so the total cost is similar to
single-head attention at full dimensionality.
Next steps
- Run
/literature-notes-query --tag attentionto see everything you have on this idea. - After a few more sources, run
/literature-notes-reviewto confront your notes with the research scope. - Run
/literature-notes-validate --check-quotesbefore you commit.literature-notes/, to confirm each quote appears on the page you cited.
Connects securely to your tools. The creator never sees your data.
About this skill
A system of 8 agent skills that turns academic PDFs (articles, dissertations, theses, books) into a structured LLM wiki of reading notes, anchored to page numbers, cited in the norm your research follows, and read against your research scope.
Why I built it
I built these skills to help me study and write my master's dissertation. Two goals shaped them:
- A second brain for academic research. Every reading becomes small, linked, searchable notes, each tied to a page and to the question I am asking. I can find what I read, and see how one author connects to another.
- Input for writing. The notes keep the quote, the page and the citation in one consistent structure, so an agent, or I, can draft from them and every claim traces back to its source. The pack does not write the paper for you. It gives you material you can trust and trace.
The aim was a system that is easy to control and to interact with: plain Markdown files that you own, a handful of commands, and a workflow you can see.
What you get
- Setup, research scope, annotate, query, review and validate, as one workflow.
- Your citation norm. Setup asks which norm you follow (APA, ABNT, MLA, Chicago or another formal norm), looks up the rules, shows you a draft and freezes it only after you approve. Every note carries its citation in that norm.
- Page handling that matches your source: cite PDF pages or printed pages, with an offset when they differ.
- Three note kinds: authorial (your own idea), interpretive (your reading of the author) and summary (a neutral account).
- A research scope (
RESEARCH.md) and a controlled glossary (Glossary.md) that every note is read against, and a review step that interviews you and updates them as your reading changes your mind. - Query by tag, source, note kind, page, influence links, verification status or free text.
- Validation of broken references (unknown tags, dangling links, undefined glossary terms), with an optional check that each quote really appears on the page you cited.
- A read-only HTML "second brain" with full-text search, backlinks and a link graph.
- Diagrams of the workflow and two complete example bundles (APA 7 on two machine learning papers, and ABNT on judicial activism) that validate cleanly.
Why it's different
The notes stay accountable to your research question. A query only loads what matches, so the bundle can grow to hundreds of notes without bloating the context. Generated prose is written in your own language, and quotes stay in the source's language.
Requirements
Python 3. The skills install pyyaml and pypdf with pip on first use. literature-notes-review also uses the free third-party grilling skill. The HTML viewer loads fonts and libraries from public CDNs.
Single-user license. Your notes are yours.
Requires Python 3 and pip. The HTML viewer needs internet access for its CDN libraries. Notes and citations are generated by a model, so verify them against the original source (the optional quote check helps). The rules for your citation norm are drafted from the sources the agent finds and only take effect after you approve them. Single-user license.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 15 days ago
Requires