Research Literature Notes
A reading-to-synthesis pipeline for papers, books, and reports. Capture the source, extract the method and the finding, link to your own take. Claude and ChatGPT do real literature work from real notes.
Requires an AI connected to your Hjarni account via MCP.
Copy this URL and paste it into Claude or ChatGPT to install the template.
How to use
- 1 Share this page. Paste this URL into Claude or ChatGPT. Your AI reads the template definition and installs it.
- 2 Folders, tags, and instructions appear. Your AI creates the full structure in your Hjarni account, ready to use.
- 3 Start adding notes. The AI instructions guide your AI on where to put things and how to organize them.
A reading-to-synthesis pipeline. Source, method, finding, your own take. Claude and ChatGPT do real literature work from real notes.
A literature review built from training data is a guess wearing a citation. This template ships the structure that keeps every synthesis grounded in a paper you actually read.
A paper note format
Every paper note follows the same shape. The starter note in the Papers folder is this skeleton.
# <First author> <Year>: <Short title>
Citation:
Abstract:
Method:
Key findings:
Limitations:
Your notes:
Related work:
Atomic findings, not paper summaries
Findings live in their own folder, one sentence per note, each linked back to the source paper. When you ask Claude or ChatGPT for evidence on a topic, it can return a list of concrete findings with authors and years, not a paraphrased blur.
Themes are slow synthesis. Findings are the raw material. Keep them separate.
A workflow that earns the template's keep
- Read a paper. Write the paper note in your own words.
- Extract one or two atomic findings into the Findings folder, linked to the paper.
- When a theme emerges across two or three papers, create a Themes note that links the supporting findings.
- When you draft, ask Claude or ChatGPT to support a claim from Findings, not from training data.
A real example
You are writing a lit review section on memory in LLM agents. You ask Claude, "What does the recent literature say about recall degrading before generation?" Claude reads the Findings folder, returns three findings with author and year, and pulls quotes from the source paper notes. You paste the synthesis, with real citations, into the draft.
Common questions
Common questions
Does Hjarni replace Zotero or Mendeley?
No. Hjarni stores your notes about sources. Bring citations and PDFs from whichever reference manager you already use.
Will the AI invent citations?
Folder instructions tell each AI to never fabricate citations and to cite by author and year only when the paper note exists. If a paper is not in the folder, the AI says so.
Can my lab share this?
Yes. Install in a team space so every researcher's Claude or ChatGPT reads from the same literature notes.
How granular should findings be?
One sentence per finding. If a sentence covers two distinct claims, split it.
Related pages
Structure
Tags
Folders
For your AI
Share this page with your AI. It reads the definition below, creates the folders, instructions, tags, and starter notes in your account.
Show template definition
Install steps for AI agents:
- Check existing tags with
tags-list. Only create missing ones withtags-create. - Create containers top-down using
containers-create, noting the returned IDs. Useparent_idto build the hierarchy. - For each container with
llm_instructions, callinstructions-updatewithlevel: "container"and the container's ID. - Create any seed notes using
notes-create, placing them in the correct container by ID. Usecontainer_pathto resolve which container. - Discuss any customizations with the user before or after installing.
---
name: Research Literature Notes
description: 'A reading-to-synthesis pipeline for papers, books, and reports. Capture
the source, extract the method and the finding, link to your own take. Claude and
ChatGPT do real literature work from real notes.
'
tags:
- paper
- method
- finding
- theme
- open-question
containers:
- name: Research Literature Notes
description: Your reading pipeline. From source, to method, to finding, to your
own synthesis.
llm_instructions: |
This is a literature notes system. The goal is to ground every synthesis in actual papers, with traceable citations.
- Never invent a paper, author, citation, finding, or quote. If the user asks for support that the folder does not contain, say so plainly.
- When summarizing literature, cite each paper by author and year. Quote findings from the Findings folder, not from training data.
- Treat the Papers folder as the source of truth. Findings, Methods, and Themes are extracted from it. If they drift, flag it.
- When the user reads something new, suggest creating both a Papers note and matching Findings or Methods notes.
- Do not paraphrase a paper's claim into a stronger version. Keep the strength of the claim as written.
children:
- name: Papers
description: One note per source. Citation, abstract, your annotated notes.
llm_instructions: |
Use this folder for raw paper notes.
- One note per source. Title format: "<First author> <Year>: <Short title>".
- Use the shipped skeleton: Citation, Abstract, Method, Key findings, Limitations, Your notes, Related work.
- Always cite by author and year when referencing a paper note. Keep direct quotes inside quote marks.
- Tag every note with "paper".
- name: Methods
description: Reusable patterns. Study designs, statistical approaches, evaluation
protocols.
llm_instructions: |
Use this folder for method patterns extracted across papers.
- One method per note. Title is a noun phrase, not a sentence.
- Cross-link to the Papers notes that use the method.
- When the user mentions a method in conversation that fits an existing note, suggest adding the new reference there.
- Tag every note with "method".
- name: Findings
description: Atomic findings, each linked to the source paper.
llm_instructions: |
Use this folder for one-sentence findings extracted from papers.
- One finding per note. Keep the sentence concrete. Avoid generic claims like "X is important".
- Always link to the source paper note with a wiki-link.
- When asked for evidence on a topic, return findings with their source by author and year.
- Tag every note with "finding".
- name: Themes
description: Cross-paper patterns and synthesis.
llm_instructions: |
Use this folder for themes that emerge from multiple papers.
- One theme per note. Title is a short noun phrase.
- List supporting Findings notes. If only one paper supports the theme, say so plainly.
- Update theme notes in place when new evidence arrives. Do not append in a journal style.
- name: Open Questions
description: Gaps in the literature you have not resolved yet.
llm_instructions: |
Use this folder for unresolved research questions.
- Each note: the question, why it matters, what you have tried, what would settle it.
- When a question is resolved by a new paper, fold the answer into the relevant Findings or Themes note and tag the question "resolved".
- Tag every note with "open-question".
notes:
- title: 'PLACEHOLDER paper: Last-name (Year), short title'
body: |
This is a placeholder paper note, not a real source. Replace the entire note with your first real paper before quoting from it.
## Citation
Last-name, F. (Year). Title of the paper. Journal name, volume(issue).
## Abstract
One paragraph summary in your own words.
## Method
Describe the study design, dataset, and evaluation protocol.
## Key findings
- Finding one, in one sentence.
- Finding two, in one sentence.
## Limitations
- Limitation one.
- Limitation two.
## Your notes
What you think this paper means for your own work. Where it agrees with or contradicts other sources you have read.
## Related work
- Link to a second paper note once you have one. Do not invent citations here.
Replace this entire placeholder before using the folder for real work.
tags:
- paper
container_path: Research Literature Notes > Papers
- title: 'PLACEHOLDER method: shape of a method note'
body: |
This is a placeholder method note, not a real method. Replace it with a real method pattern before using the folder for real work.
## What it is
One or two sentences describing the method in plain language.
## When to use
The kinds of questions this method is built to answer.
## Caveats
The conditions under which the method fails or misleads.
## Used by
- Link to the paper notes that use this method once you have them. Do not link to invented papers.
Replace this entire placeholder before using the folder for real work.
tags:
- method
container_path: Research Literature Notes > Methods
- title: 'PLACEHOLDER finding: shape of a finding note'
body: |
A finding lives as one concrete sentence, linked to the paper it came from.
Source: link to a real paper note once you have one. Do not source findings to invented papers.
Replace this entire placeholder before using the folder for real work.
tags:
- finding
container_path: Research Literature Notes > Findings
- title: 'PLACEHOLDER theme: shape of a theme note'
body: |
A theme lives as a short noun phrase title and a paragraph that ties together two or more real findings.
## Summary
One paragraph naming the pattern across the supporting findings.
## Supporting findings
- Link to real finding notes once you have them. Two or more is the bar for a theme.
## Open questions
- What would need to be true for the theme to break?
Replace this entire placeholder before using the folder for real work.
This is a starter note. Replace it with a real theme.
tags:
- finding
container_path: Research Literature Notes > Themes
- title: 'Open question: How to evaluate memory faithfulness'
body: |
## Question
What is the right protocol for measuring whether a memory-augmented agent stays faithful to its retrieved notes?
## Why it matters
Faithfulness is the property that makes the rest of the system worth shipping.
## What we know
- Standard QA benchmarks do not capture this directly.
## What we do not know
- Whether a single faithfulness score is meaningful, or whether the metric needs to be split by task type.
This is a starter note. Replace it with a real open question.
tags:
- open-question
container_path: Research Literature Notes > Open Questions