wiki-checkout
Skill1 starsChanged 25 days agoMITView on GitHub
The librarian: choose the wiki documents that serve a request, follow their links only while the documents still serve it, and check out copies with a reading list in reading order. Return proposes the user's edits of the copies as one change. Use for check out the material on X, pull together what we know about X, make me a reading stack on X, a reading list on X, return my checkout. An answer to a question is wiki-query.
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 2 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 24
Call checkout with action: candidates, text set to the request, and the tags.
- Very specific: line 25
Add types: [topic, repository, source] when the user asks for the sources themselves ("the papers on X").
- Distinctive: line 37
Follow a branch past the list only while it serves.
- Very specific: line 48
Call checkout with action: make, and:
- Distinctive: line 63
When the user asks to return a checkout:
- Very specific: line 68
Call checkout with action: return and folder.
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 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
wiki-checkout
A checkout is a folder of checkout/ that holds copies of wiki documents and a reading
list. The user reads and edits the copies in Obsidian. Code ranks the candidates and
writes the folder. This skill chooses the documents, puts them in reading order, and
gives each one a reason. Return turns the edited copies into one proposed change of
their originals.
Tools: checkout (candidates, make, list, return), and Read. References:
changes.md.
Procedure
- Take the request in the user's words. The palette sends
/almagest:wiki-checkout Check out the material on: <request>; the text after the colon is the request. When the request names a project or a category, turn the words that name tags of the vocabulary intotags(the tag list in the session-start context, orvault). - Call
checkoutwithaction: candidates,textset to the request, and thetags. Addtypes: [topic, repository, source]when the user asks for the sources themselves ("the papers on X"). Code ranks every document with the search, takes the best hits as seeds, and adds the documents linked to or from them, up to two links away. Each candidate hasref(with its description),score,distance(0 for a search hit),via(the title it was reached from), andwords(the length of its body). - Choose. Read each candidate's description. When the description does not settle whether the document serves the request, read the document with Read. Keep a document that serves the request; drop one that does not. Judge a linked document on its own content, not on the hit that led to it. When you drop a document, also drop the candidates reached only through it, unless they serve the request on their own. So a branch stops where its documents stop serving.
- Follow a branch past the list only while it serves. The candidates stop two links from a search hit. When a kept document links others that serve the request, read them and keep the ones that do.
- Size the checkout for a sitting or a week of reading: at most about 30 documents.
wordstells you how long each one is. When more than about 30 documents serve, the request is too broad: ask the user once to narrow it, and offer two or three narrower requests from the tags or subjects of the candidates. After the answer, make the checkout. - Order the documents foundations first: a definition, an overview, or a concept that
the others assume comes before the documents that build on it. Give each document a
why: one line that says what it gives the reader for this request. Call
checkoutwithaction: make, and:request: the user's words;name: a short name for the folder, such as "Reinforcement learning";documents:[{id, why}]in reading order;notes: what you left out and why: the near subjects you dropped, the branches you stopped, and a gap the wiki has.
Code writes
checkout/<date> <name>/in one commit: a copy of each document as<Title> (checkout).md, and the checkout's index_index.md(its name, request, status, and reading order). The ledgercheckout/Checkout · Ledger.mdis a Base of every checkout's index. A link between two documents of the checkout points at the copy; every other link points at the wiki.
Return
When the user asks to return a checkout:
- Find its folder. When the user does not name it, call
checkout(action: list) and take the checkout withstatus: outwhose request matches. When two match, ask which, and name both. - Call
checkoutwithaction: returnandfolder. Code makes onemodifywrite for each edited copy, with the copy's text and its links pointed back at the originals, and proposes the change "Return<name>". Then it moves the checkout, every file of it, totool/returned/<folder>/, marks its index returned, and keeps it there as the user left it. Return carries a copy's body only; an edit of a copy's frontmatter does not return. - Show the result in one or two lines: the change document as a link, when there is
one, the checkout's place now, and each copy in
skippedwith its reason. A copy is skipped when its original is gone or changed since the checkout; its edits stay in the returned copy. With no edited copy, there is no change: the checkout just moves. - A checkout returns once. When the tool says it was returned already, tell the user to check the documents out again to work on them further.
Gate
None for make: a checkout changes no knowledge, and the copies are the user's. Never
edit a file in checkout/ or tool/returned/; the guard refuses it. Return proposes a change, and the user
decides in the change document (Approve or Cancel). Never apply it: the change names no
session, so the change tool refuses your apply.
End with one or two lines that name the checkout's index as a link
([[checkout/<folder>/_index|<name>]], as make returns it in index) and the count of
documents.
Hand off
wiki-query for a question about the material. wiki-edit to carry the edits of a skipped copy into the wiki, or for any other change to the originals.
Other versions2 projects use a version of this.
Word for word · 2
- nathanaday/almagest★ 1100% same
- nathanaday/claude-atlas★ 1100% same
Edited · 0
None found.
Wrote their own version · 0
None found.
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository15 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- almagest-onboardSkill
- almagestSkill
- repo-ingestSkill
- repo-linkSkill
- repo-unlinkSkill
- wiki-editSkill
- wiki-ingestSkill
- wiki-mapSkill
- wiki-querySkill
- wiki-reviewSkill
- wiki-saveSkill
- wiki-syncSkill
- wiki-wikifySkill
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.

