gtd-clarify
Skill0 starsChanged 5 months agoMITView on GitHub
Process the GTD inbox — turn raw captured items into decided outcomes and next actions. Use when the user wants to sort out / organize / process their inbox, "整理する", "クラリファイ", "これどうしよう", or when 未処理 items are piling up. Reads each item WITH UNDERSTANDING (including the user's casual hand-written notes), asks the two magic questions, routes each to do·delegate·defer / someday / reference / trash / calendar, and records the result in Cosense. Part of the gtd skill bundle.
In this repository4 other files it gives its agents.
- gtd-captureSkill
- gtd-engageSkill
- gtd-reviewSkill
- gtdSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 1 is distinctive, and 4 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.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 21
Start by actually reading the inbox — python3 "$GTD" show (or browsePage for full text).
- Distinctive: line 29
Don't skip to the easy one, and don't leave an item in limbo — that limbo is what nags.
- Very specific: line 54
Prefer a project title that is the done-state ([契約をクラウド化する], [確定申告を提出する]) — then the name itself says what "done" looks like and you can …
- Very specific: line 73
If an item is tied to a specific day/time ("7/20のイベント登壇", "金曜15時に歯医者"), it belongs on the calendar, not as an inbox next:.
- Very specific: line 86
Use the script — it finds items by fuzzy text and puts things in the right place.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
gtd-clarify — get "in" to empty by deciding, not doing
Clarifying is deciding what each thing is and what to do about it — not doing the work.
"In to empty" = every item in 📥 未処理 has been decided (then it moves to 🗂 整理済み).
GTD=~/.claude/skills/gtd/scripts/gtd_cosense.py.
Read first, with understanding
Start by actually reading the inbox — python3 "$GTD" show (or browsePage for full text).
Take in what the user wrote, including their free-form notes, arrows, and half-thoughts —
those are clarify input, not clutter. If the user hand-wrote "A社よさげ→来週決めたい" under an
item, that is the direction; fold it in rather than asking from scratch. The 📥 未処理
items are what need deciding; already-clear lines in 🗂 整理済み can be left alone.
Work one item at a time, top to bottom
Don't skip to the easy one, and don't leave an item in limbo — that limbo is what nags. Ask one good question at a time, not a questionnaire. For each item:
What is it? (read the user's words) → Is it actionable?
├─ NO → trash → complete --match (just remove it)
│ someday/maybe → move-someday --match
│ reference → note it onto a page, then complete --match
└─ YES → What's the NEXT physical, visible action?
├─ <2 min? → do it now, then complete --match
├─ day/time-fixed? → Google Calendar (below), not a next:
├─ someone else? → set-fields --waiting "相手・何を (依頼: YYYY-MM-DD)"
└─ else → set-fields --next "最初の物理的な一歩" (this promotes it to 🗂)
Will that action finish the whole thing?
├─ NO → it's a project → name it as its OUTCOME and give it a page:
│ bracketize --match … --title "契約をクラウド化する" --page
└─ YES → a bare next-action line is fineThe pivot is "what's the next action?", and it must be a concrete physical act. "会議を 設定する" / "決める" aren't actions yet — ask 「具体的には最初に何をする?誰に、どうやって?」 until you reach call whom / email whom / draft what / open which page. Better wrong than vague.
Name projects as outcomes; goal only when needed
Prefer a project title that is the done-state ([契約をクラウド化する], [確定申告を提出する])
— then the name itself says what "done" looks like and you can skip goal:. Add a goal: line
only when the title doesn't make the outcome obvious. This is lower-friction and truer to Allen
(he names projects as outcomes).
Questions to ask (Japanese; pick the one that unblocks the item)
- Identify / actionable: 「これ何?」「何かやることある?それとも情報/いつか?」
- Outcome (when they don't know what they want): 「どうなったら“完了”って言える?」
- Next action (when they know the outcome but are stuck): 「次に取る、具体的な最初の一歩は? もし今これだけやれるとしたら、まず何をする?」
- Delegate: 「自分がやる?誰かに渡す?」→ if handoff: 「いつ頼んだことにする(日付)?」
- Someday: 「今やる?“いつかやるかも”に置いておく?」
Snyder's split when stuck: don't know what they want → work the outcome ("make it up"); know it but frozen → extract the next action ("make it happen").
Hard landscape → Google Calendar
If an item is tied to a specific day/time ("7/20のイベント登壇", "金曜15時に歯医者"), it belongs
on the calendar, not as an inbox next:. Use the connected Google Calendar MCP — the tools
are create_event (and list_events / list_calendars); if they're not already loaded, find
them with ToolSearch (query: calendar). Confirm the date/time/title with the user, create the
event, then handle any non-dated remainder (prep work) as a normal next:.
create_event(summary="○○イベントで登壇", startTime="2026-07-20T13:00:00",
endTime="2026-07-20T14:00:00", timeZone="Asia/Tokyo")Recording it (dry-run, then commit)
Use the script — it finds items by fuzzy text and puts things in the right place. Run without
--commit first when the change isn't obvious; show the user; then add --commit. For thinking
/ logs / anything free-form, use note (any wording, no schema):
python3 "$GTD" set-fields --match 電子契約 --next "候補3社の料金ページを開いて比較表を作る" --commit
python3 "$GTD" bracketize --match 電子契約 --title "契約をクラウド化する" --page --commit <<'EOF'
from [inbox]
EOF
python3 "$GTD" note --match 契約 --text "A社よさげ、B社は高い。来週決める" --commit
python3 "$GTD" move-someday --match 英語 --commit
python3 "$GTD" complete --match 牛乳 --commitTone
Be a calm coach, not a form to fill in. Confirm each decision briefly (「next 書いた」) and move
on. When 📥 未処理 is empty, say so — that's "in to empty", and it's the moment the mind lets go.
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.

