collate
Skill9 starsChanged 5 months agoMITView on GitHub
Collate personal notes and sources into a paper or concept page, or update an existing page with a new source
In this repository14 other files it gives its agents.
- .claude/AGENTS.mdAGENTS.md
- .claude/CLAUDE.mdCLAUDE.md
- alphaxivSkill
- askSkill
- autoexploreSkill
- curriculumSkill
- editSkill
- goalSkill
- notionSkill
- obsidianSkill
- retrieveSkill
- tutorSkill
- webSkill
- zoteroSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
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 11
If $ARGUMENTS is empty, ask the user for a topic or paper name, then continue.
- Very specific: line 28
Run the BM25 retriever to get seed notes, then expand by following relevant wikilinks — exactly as the /retrieve skill does:
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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
Collate all personal notes and source material for $ARGUMENTS into a single paper or concept page, distinguishing personal annotations from objective content. Also handles updating existing pages when a new source is added.
If $ARGUMENTS is empty, ask the user for a topic or paper name, then continue.
Step 0 — Check ingestion
Glob sources/ and notes/ for any files matching $ARGUMENTS (by slug or
title). If nothing is found, tell the user and prompt them to run the appropriate
ingestion skill first:
- PDF →
/zotero <path> - Notion export →
/notion <path> - Obsidian note →
/obsidian <path> - arXiv paper →
/alphaxiv <id or title>
Then stop. If files are found, continue to Step 1.
Step 1 — Retrieve
Run the BM25 retriever to get seed notes, then expand by following relevant
wikilinks — exactly as the /retrieve skill does:
Each output line issh .claude/src/pyrun .claude/src/retrieve_cli.py --retrieve "$ARGUMENTS"<file path>\t<title>. These are your seed notes.- For each seed,
Readit and follow any[[wikilinks]]you judge relevant to $ARGUMENTS, resolving them with:
Output issh .claude/src/pyrun .claude/src/retrieve_cli.py --resolve "<link>"<link>\t<path|NONE>; skipNONE. Stop after ~1–2 hops.
The relevant set covers all four dirs: sources/, notes/, papers/, concepts/.
If retrieve returns nothing relevant, name the closest topics already covered in the base and stop without writing. Do not attempt to write a page with no source material.
Step 2 — Detect type
Decide: is $ARGUMENTS a paper (specific published work) or a concept (topic, idea, technique)?
- If a paper → output to
papers/<slug>.md, use paper template - If a concept → output to
concepts/<slug>.md, use concept template
If unsure, ask the user.
Step 3 — Check for existing page
Slugify $ARGUMENTS to kebab-case and check whether papers/<slug>.md or
concepts/<slug>.md already exists.
- Existing page found → read it and ask the user: "I found an existing
page for
<title>— is this the one you want to update, or are you working on a different sense of the term?" If it's the same, switch to Update Mode (see below). If different, ask for a clarifying name to disambiguate the slug, then continue to Step 4. - No existing page → continue to Step 4 (create from scratch).
Update Mode — adding a new source to an existing page
Use this when a page already exists and a new source has been ingested that is relevant to it.
U1 — Identify new sources
Read the existing page's sources: frontmatter list. Compare against all
retrieved files from Step 1. Any file not already listed is a new source.
If there are no new sources, tell the user the page is already up to date and stop.
U2 — Merge personal notes and annotations
For each new source:
- If from
notes/notion_*.mdornotes/obsidian_*.md: append to## My Notesas a new paragraph, clearly separated from existing content. - If from
notes/zotero_*/annotations.md: weave new annotations into the relevant objective content sections as callouts, following the annotation rendering rules in Step 5. Never duplicate an annotation already present.
U3 — Merge objective content
Read the new source's ## Content. Identify any claims, findings, or details
not already covered in the existing page's ## Summary sections.
Add only genuinely new information — do not rewrite existing content.
Append to the relevant section or add a new section if needed.
Cite inline with a relative link to the new source file.
U4 — Update frontmatter
Add the new source path to sources: in frontmatter. Update created → keep
original; do not change it.
U5 — Write
Overwrite the existing file with the merged content. Report what was added.
Step 4 — My Notes section (new page)
Read the relevant template (.claude/Templates/paper.md or .claude/Templates/concept.md).
Gather Notion/Obsidian material (files matching notes/notion_*.md or
notes/obsidian_*.md) into a ## My Notes section at the top of the page:
- Include the full note content as plain paragraphs, preserving the author's words.
- Preserve all
[[wikilinks]]and#tagsexactly as written — do not convert or strip them. - If there are no Notion/Obsidian notes, omit the section entirely.
Step 5 — Objective Summary with Annotations (new page)
Write the objective summary from sources/ material (Zotero source files,
AlphaXiv sources). Follow the appropriate template structure:
Paper: Summary, Key Contributions, Methodology, Important Concepts, Connections to Related Work, Limitations & Open Questions.
Concept: Overview, thematic sections, Related Concepts.
Follow the Page conventions in .claude/AGENTS.md — grounding & citations,
mathematics in LaTeX (be mathematical where the material is; include critical
equations and derivations), and cross-linking. Read .claude/AGENTS.md if
those conventions aren't already in context.
While writing each section, check the Zotero annotation file(s) (files under
notes/zotero_*/annotations.md) for annotations that touch the content of that
section. Use the annotation's position description and annotated text to determine
where it belongs. Weave the annotation in immediately after the relevant
paragraph using the following rules:
Annotation rendering rules
Ink underline — no question or comment alongside:
Bold the underlined phrase/sentence inline within the objective prose. The bold
signals the reader found this important without adding visual clutter.
Example: the sentence containing "the Transformer is the first transduction model relying entirely on self-attention" becomes **the Transformer is the first transduction model relying entirely on self-attention**.
Ink underline — with a question or comment written alongside:
Bold the underlined text in the prose, then immediately after the paragraph add
a [!question] callout quoting the question:
> [!question]
> "question text as written"Handwritten question (standalone, not tied to an underline):
Add a [!question] callout immediately after the paragraph most relevant to
that question:
> [!question]
> "question text as written"Handwritten comment or expansion (not a question):
Add a [!note] callout immediately after the relevant paragraph:
> [!note]
> "comment text as written"Ink circles, strokes, or marks on figures/diagrams:
Add a [!note] callout near the prose describing that figure or diagram, briefly
noting what was circled or marked and what it likely indicates:
> [!note]
> Circled [element] in Figure N — marks [what the reader was tracking].Skip any annotation that is too vague to place (e.g. a single mark with no discernible text or figure reference), and note it in the wrap-up.
Step 6 — Fill frontmatter (new page)
Paper frontmatter:
title, type: paper, created, authors, year, related_papers, related_concepts, sourcesConcept frontmatter:
title, type: concept, created, related_concepts, related_papers, sourcesauthors and year are paper-only. Extract from the source material if available.
Step 7 — Write (new page)
Write to papers/<slug>.md or concepts/<slug>.md (slug from Step 3).
Wrap up
Report: file written or updated, which sources/ and notes/ files were used,
anything skipped (one-line reason each). Then ask the user if they're happy or
want to refine — loop via /edit <page path> until they are. Do not reindex
manually — the background poller rebuilds the index automatically after edits settle.
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.

