improve-cards
Skill3 starsChanged 19 days agoMITView on GitHub
Works on the user's existing Anki cards - searches the collection, finds the cards they struggle with from their review history (again rate, lapses, FSRS difficulty, leeches, red flags), works out why each one fails and fixes it without losing its review history, deletes cards and undoes earlier changes. Use when the user asks what is in their collection, wants cards corrected, improved, cleaned up or removed, asks which cards are hard for them, or wants to undo something this plugin did.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- MichaelPrutkow/anki-for-claudeword for word, first seen Oct 17, 2026
Other versions1 project uses a version of this.
Word for word · 1
- MichaelPrutkow/anki-for-claude★ 3100% 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 repository4 other files it gives its agents.
- check-setupSkill
- create-cardsSkill
- exam-prepSkill
- visualsSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 1 is distinctive, and 3 are 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.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 37
No hits usually means the query is malformed, not that nothing matches.
- Very specific: line 41
Call sync, so reviews from the phone count.
- Very specific: line 42
Call stats, with scope set to the deck the user cares about (for example deck:"Analysis 2").
- Very specific: line 67
Delete a card only when it should not exist at all, and ask the user first.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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
Improving existing cards
If overview and the other anki tools are not among your tools, call get_device_info (search for it if it is not loaded) and search for the anki tools again. In the Claude app, a chat reaches the tools on the user's computer reliably only if the user added a folder to it before sending the first message. If the tools are still missing, ask the user to start a new chat, add a folder with the Add folder button (for example the folder with their study material) and send the request again. Only if that does not help either, follow the check-setup skill.
Finding cards
search_notes takes Anki's search syntax, and the query has to follow it exactly:
| Syntax | Meaning |
dog cat | both words (a space means AND) |
dog or cat | either word |
-cat | without cat |
dog (cat or mouse) | grouping |
"a dog" | exact phrase; also needed for spaces in names: deck:"French Words" |
d*g | wildcard: dog, dug, dg |
deck:Name, deck:Name::Sub | a deck, including its subdecks |
tag:Name, tag:none | a tag; tags are hierarchical, tag:mcp::batch::* matches every batch |
note:Basic | a note type |
Front:dog | the whole field is exactly "dog"; Front:*dog* finds it anywhere in the field |
Front: / Front:_* | field empty / not empty |
nid:123,456 | notes by id |
is:new, is:due, is:learn, is:review, is:suspended | card state |
prop:lapses>3, prop:ivl>=21 | numeric card properties |
rated:7:1 | answered "again" in the last 7 days |
added:7, edited:7 | added or edited in the last 7 days |
flag:1 | red flag |
\*, \_, \", \: | a literal *, _, " or : |
Cards created by this plugin have the tag mcp::created, and each commit adds mcp::batch::<id>.
No hits usually means the query is malformed, not that nothing matches.
Finding the cards that do not work
- Call
sync, so reviews from the phone count. If it fails, go on without it. - Call
stats, withscopeset to the deck the user cares about (for exampledeck:"Analysis 2"). - For each weak note, read its fields (
search_noteswithnid:<id>and a highermax_charsshows all of it) and pick the most likely cause:
| Signal | Likely cause | Fix |
| LONG_ANSWER, LIST_LIKE, MANY_CLOZES | the card asks for several facts | split it: rewrite the card into the first part and add the others as new cards |
| high again rate, short answer | the question is ambiguous, or a similar card interferes | add the qualifier that pins the answer down |
| SLOW | the answer is worked out, not recalled | add an example card or a picture (visuals skill) |
| LEECH, many lapses | the concept is not understood | add a prerequisite card or a small worked example |
| FLAGGED | the user marked it while studying | the strongest signal; read the card carefully |
| none of the above | the material is hard | leave it and say so |
- Propose the fixes in the chat, note by note: what is wrong and the new text. New cards from splits, examples and prerequisites go into one draft with
draft_add, written by the rules of thecreate-cardsskill. - Change nothing until the user approves. Then apply the approved rewrites with
update_notes, add the approved new cards withdraft_commit, and report the batch ids.
Changing cards
update_noteschanges fields and keeps the review history. Prefer it to deleting and adding a new card.- Send only the fields that change, each with its full new content.
- Notes without the tag
mcp::createdwere written by the user. Ask before changing them, then callupdate_notesagain withforce=true. - The formatting rules of the
create-cardsskill apply to every field you write. - To add a picture to an existing card, make it with the tools from the
visualsskill and put its<img>tag into the field.
Deleting
Delete a card only when it should not exist at all, and ask the user first. delete_notes removes only notes this plugin created, takes their review history with them and cannot be undone.
Undoing
draft_commitandupdate_notesreturn a batch id.undowith that id deletes the notes a commit added, or restores the fields an update changed.- A note the user edited since is left alone and reported.
- Deleting cannot be undone.
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.

