logseq-paper-notes

---

name: logseq-paper-notes description: Create and update paper reading notes in the user's Logseq vault. Use when summarizing a paper, recording results from a paper reading session, or appending notes to an existing paper page. The vault is shared between Logseq (daily-use viewer) and Obsidian (CLI access). Writes use Logseq syntax (key:: value, tab-indented outline), reads can use Obsidian CLI for backlinks/search or rg for raw content.

Logseq Paper Notes

Manage paper notes in the user's Logseq vault using a fuzzy convention: minimal file naming/property rules, rg for search, and Obsidian CLI for graph operations (backlinks, tag listing, search).

Vault location

• Pages directory: $LOGSEQ_PAGES_ROOT (resolve via fish -i -c 'echo $LOGSEQ_PAGES_ROOT' if env var isn't visible from bash)
• Layout: flat — all pages live directly under the pages directory, no subdirectories
• Vault is also registered in Obsidian as vault name pages with "cli":true

File naming

Two categories of paper notes coexist in the vault:

1. Zotero-imported notes — filenames start with @ (e.g. @A survey on the complexity of learning quantum states.md).

   • NEVER create new files with @ prefix. This prefix is reserved for Zotero importer output.
   • When the agent finds an existing @-file matching a paper, append to it instead of creating a new file.

2. Agent-created notes — choose filename in this order:

   • If arxiv ID is known: {arxiv_id}.md (e.g. 2401.12345.md)
   • If DOI is known but no arxiv ID: {doi-slug}.md with / → _
   • Otherwise: short English-title .md (no @ prefix)
   • URL-encode : → %3A and other filesystem-unsafe chars (matching the Zotero convention used in existing notes)

Property format (Logseq syntax)

Logseq page properties are key:: value lines (NOT YAML frontmatter) and must be the first block of the page to be recognized as page properties.

When creating a new agent-managed note, write only properties the agent can confidently fill:

tags:: [[Tag1]], [[Tag2]]
authors:: [[Firstname Lastname]], [[Another Author]]
date:: YYYY-MM-DD
item-type:: [[preprint]]
title:: Paper Full Title
arxiv-id:: 2401.XXXXX
doi:: 10.XXXX/YYYY
url:: https://arxiv.org/abs/2401.XXXXX

Notes:

• item-type values seen in existing notes: [[preprint]], [[journalArticle]]. Match those.
• tags:: and authors:: use [[page link]] syntax with comma separation.
• Do NOT include Zotero-specific properties (library-catalog, access-date, zotero:// links, extra::) — those are meaningful only for Zotero-imported notes.

After the property block, use Logseq outline structure with tab indentation:

- [[Abstract]]
	- Paper abstract here.
- [[Notes]]
	- Freeform reading notes — summary, key results, and comments all go here as sub-blocks.

Which [[...]] section headers are real, and why they are links. Only three section headers actually recur in the vault and aggregate across the graph: [[Abstract]] (~600 notes), [[Attachments]] (~675 — Zotero-machine-generated, carries {{zotero-imported-file ...}} macros; present when appending to a @ note, do NOT fabricate it in agent-created notes), and [[Notes]] (~125). They are wikilinks on purpose: opening [[Abstract]] collects every paper's abstract via linked references (Logseq materializes the page from the references even with no Abstract.md file). Do NOT invent sparse sections — earlier versions of this skill listed [[Summary]] (2 uses) and [[Key results]] (0 uses), which aggregate nothing and just create near-orphan pages; fold that content under [[Notes]]. This link-as-section idiom is specific to paper notes: for non-paper or uncurated notes (e.g. a brainstorm log), use plain text or **bold** headers, because every [[...]] spawns a page (junk like a Seed/Status page). Reserve [[...]] for nodes you actually want in the graph.

Discover existing tag vocabulary before adding new tags

Before inventing a new tag, check what's already used:

rg -oN --no-filename '^tags::.*' "$LOGSEQ_PAGES_ROOT" \
  | rg -o '\[\[[^\]]+\]\]' | sort -u

Match existing capitalization and structure (e.g. [[Quantum Physics]], [[Computer Science - Machine Learning]]).

Searching/querying notes

Find a paper by arxiv ID, DOI, or partial title (use rg)

rg -l 'arxiv-id:: 2401' "$LOGSEQ_PAGES_ROOT"
rg -l 'doi:: 10.1103' "$LOGSEQ_PAGES_ROOT"
fd -i 'partial title' "$LOGSEQ_PAGES_ROOT"

List notes with a given tag

rg -l 'tags::.*\[\[Tensor Network' "$LOGSEQ_PAGES_ROOT"

Backlinks / link graph (use Obsidian CLI)

Obsidian CLI gives graph-aware operations rg cannot:

obsidian backlinks file="@Tensor Network Representations of Parton Wave Functions" format=json
obsidian links file=2401.12345
obsidian unresolved
obsidian orphans
obsidian tags

Obsidian CLI requires the GUI to be running with the pages vault open. If commands return Vault not found., launch with open -a Obsidian and wait ~5s before retrying.

Full-text search

obsidian search query="VQ-VAE" format=json
# or via rg
rg -l 'VQ-VAE' "$LOGSEQ_PAGES_ROOT"

Writing notes

Creating a new note

Use the Write tool to write Logseq-format content directly. This is the most predictable path and avoids any CLI quirks. Place the file at $LOGSEQ_PAGES_ROOT/{filename}.md.

Alternative: obsidian create path={filename}.md content='...' — equivalent, with \n/\t escape sequences. Useful only when running other Obsidian CLI ops in the same session and wanting Obsidian's index updated immediately.

Appending to an existing note (especially Zotero @ notes)

Use the Edit tool to append a new outline block (- [[Notes]]\n\t- ...) at the end of the file. Preserve the existing outline structure and indentation (tabs).

Alternative: obsidian append path={filename}.md content='- [[Notes]]\n\t- text' — adds a blank line then the content at end of file.

Prohibited: obsidian property:set / property:remove on Logseq-managed files

These commands inject YAML frontmatter (---\nkey: value\n---) at the top of the file. Logseq requires page properties to be in the first block, so the frontmatter pushes existing key:: value lines out of property position and they stop being recognized as page properties in Logseq. Do not use these commands on this vault.

To update properties, edit the key:: value lines directly with the Edit tool.

Quick reference: Obsidian CLI parameter quirks

• property:set uses name= (not key=)
• search uses query= (not positional)
• vault=<name> defaults to active vault if omitted
• Many commands support format=json|tsv|csv
• file=<name> resolves like wikilinks (basename match), path=<path> is exact relative path
• delete moves to trash (no visible .trash directory in vault)

Run obsidian help for the full command list (~100 commands, requires GUI running).