kb-note
Skill1 starsChanged 25 days agoMITView on GitHub
Saves a small piece of information from the conversation into the user's knowledge base as a quick note, to be filed properly later by kb-ingest. Use when the user says "note this", "save this", "add this to my knowledge base", "jot this down" or similar, for example "note this: send the draft plan to Alex by Friday". Also saves a knowledge-base answer the user wants kept ("save this answer", "archive this answer") as an answer snapshot. Writes the user's words, key points and context to one new file in the library's quick-notes folder, replies with the file name and the key point, and when notes pile up suggests saying "file my quick notes". Only creates files, never changes existing ones. Use only when the user asks; never save on your own or ask whether to save. Requests aimed at the assistant's own memory ("remember that", "from now on") are not for this skill. Filing, archiving and health checks belong to kb-ingest; questions belong to kb-query.
2 other projects use a version of this, 1 of them edited.
Adopted by
Most recently taken up by
- gabrielbchacon/ai-dev-bootstraptheir own version, first seen Oct 20, 2026
- walkwithedi/evergreen-wikiedited, first seen Oct 11, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 1
- walkwithedi/evergreen-wiki★ 222% same
Wrote their own version · 1
- gabrielbchacon/ai-dev-bootstrap★ 0same name
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.
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 4 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 12
Read references/config.md and references/environment.md once before starting.
- Distinctive: line 16
If the request carries content, save that.
- Distinctive: line 17
If it is only the command ("note this"), save the most recent information the user gave or confirmed.
- Distinctive: line 18
If it could refer to several places, ask in one sentence which one.
- Very specific: line 30
After the note is saved, check that library for files waiting for consolidation (references/config.md, section 5), and remind once per …
- Distinctive: line 70
Only when write_new in step 4 succeeded and a find_exact of the file name in the notes folder finds it.
≈ “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
Quick notes: save a piece of the conversation
Version: 2.1.0
This skill does one thing: create one new file in a library's quick-notes folder. It never edits, moves or deletes any file, and never touches the library root, archive, entities or inbox.
Read references/config.md and references/environment.md once before starting.
Step 1: what to save
- If the request carries content, save that.
- If it is only the command ("note this"), save the most recent information the user gave or confirmed. If it includes your own answer, save only what the user accepted.
- If it could refer to several places, ask in one sentence which one. Do not guess.
- Unrelated things in one request: one file each.
- Tentative or unconfirmed statements are saved as they are, marked "unconfirmed" in the key points.
- Answer snapshot: when the user asks to keep a knowledge-base answer ("save this answer", "archive this answer", 「把这个回答存下来」「存档这个回答」), save the question and that answer exactly as given, with its basis lines. It is saved as a quick note like any other (step 3, with the changes listed there); it goes into the library only when the quick notes are filed.
Step 2: which library
Load the registry. Match the content against each library's scope.
- The user named a library: use it.
- Exactly one scope fits: use it.
- None or several fit: ask "Which library:
<labels>?" in one sentence. Never default. Then read that library's backend file (references/backend-gdrive.md) and its _kb.json (for language, timezone, sensitivity and names). After the note is saved, check that library for files waiting for consolidation (references/config.md, section 5), and remind once per conversation; this skill only reminds and hands "consolidate" to kb-ingest.
Step 3: write the note
Plain text inside the file, no Markdown, in this fixed layout (labels in the library language):
Quick note
Time: YYYY-MM-DD HH:MM (<timezone>)
Library: <label>
Suggested partition and topic: <partition topic> / not judged
Source: assistant conversation
Words: <the user's words, exactly; include needed context lines marked "<user_label>:" and "<assistant_label>:">
Key points:
- <one point per line, only what the words say, nothing inferred or added>
Context: <one to three sentences: what conversation or question this came up in, so it makes sense later>
Rules:
- Suggested partition and topic: only if this conversation already read the library spec's Table A; otherwise "not judged". Do not fetch the spec just for this. It is only a hint.
- With sensitivity enabled: mask ID numbers, card and account numbers, passwords and one-time codes with asterisks, keeping at most the last keep_last characters, in words and key points alike. Never add health, money or investment advice or interpretation.
- Extra hard rules in the library's _kb-params.md also apply here.
- Dates, numbers, names and organisations exactly as the user wrote them. A relative date ("next Wednesday") gets the computed date in brackets in the key points, marked "computed".
- Answer snapshot (step 1, rule 6): Source is "knowledge base answer" (「知识库问答」). Words hold the user's question and the answer exactly as given, basis lines included. The first key point is "answer snapshot of YYYY-MM-DD, not updated when its sources change" (「YYYY-MM-DD 的问答快照,不随来源更新」); the others list the basis files the answer names. The short topic in the file name starts with "answer" (「问答」).
Step 4: save it
- Find the notes folder (names.notes_dir) in the library root with find_folder.
- One: use it.
- Several (Drive): use the oldest and mention the duplicates in the reply.
- None: ensure_folder, and mention "notes folder created" in the reply.
- File name: YYYYMMDD-HHMM_
<short-topic>.md in the library timezone. The short topic is a few words summing up the content, no punctuation or slashes; in a family library include who it concerns. Several notes in the same minute: add -2, -3. - write_new in the notes folder with the content from step 3.
- On failure retry once. If it still fails, give the reason and show the full note in the reply so the user can save it or ask again later.
Step 5: confirm
Only when write_new in step 4 succeeded and a find_exact of the file name in the notes folder finds it; until then never say the note was saved. Reply in one or two lines:
Saved to <label> quick notes: <file name>
Key point: <one sentence>
Then check the backlog: list the notes folder (one level, paging until done), count .md files and find the oldest. More than 20, or the oldest older than 14 days: add a line "N quick notes waiting (oldest YYYY-MM-DD); say 'file my quick notes' to file them." Otherwise say nothing about it.
Replies follow the library language and preferences.output.
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.

