Inbox triage
Open your laptop. Your inbox is already sorted.
What you type
Run inbox triage on these:
Type it into Claude or ChatGPT once the skill is switched on. The wording does not have to be exact.
- Category
- Productivity
- Licence
- MIT
- Works with
- Gmail · Outlook
What it does.
Free Claude skill from OpenKit. Reads your morning inbox and sorts it into urgent, watching and skip. Read-only, so nothing is sent.
Once the skill is in your AI tool there is nothing to configure. You hand it the morning's emails and ask for the triage. It reads them and sorts them into urgent, watching and skip. When you want a reply, it hands off to your drafter, and nothing is ever sent on your behalf.
How does the inbox triage skill sort my email?
It reads your morning inbox and sorts it into three piles: urgent for what needs you today, watching for what you are waiting on, and skip for the newsletters and noise it counts but never opens. It works with the Gmail or Outlook content you give it, and it is read-only.
Installing it.
The skill is a folder of plain instructions your AI reads, so there is nothing to configure once it is switched on.
- 01 Download the folder
It is 8 KB of plain text and MIT-licensed, so you can read every instruction in it before you run anything.
- 02 Switch it on in your AI’s settings
Add it under Skills and turn it on. It works the same in Claude and ChatGPT, and you do not have to write any code.
- 03 Or install it from the command line
One line in a terminal does the same job:
npx @openkit/skills install inbox-triage
Is the inbox triage skill free, and do I need a paid Claude plan?
It is free and MIT-licensed on GitHub, and you do not need a paid Claude plan. Skills work on the free plan as well as Pro, Max, Team and Enterprise; you just enable code execution in your Claude settings first.
How do I install the inbox triage skill without coding?
Download it from GitHub, then upload it in your AI's settings under Skills and switch it on. No coding needed, and it works the same in Claude and ChatGPT.
Running it.
Plain instructions your AI follows. There is nothing to code.
- 01 Grab the morning's emails
In Gmail or Outlook, copy your overnight and unread messages, or forward the batch to yourself so the text is easy to paste.
- 02 Ask your AI to run it
Paste the emails into Claude and say "Run inbox triage on these." The skill activates on its own, with no command to remember.
- 03 Read the brief
You get three piles back: urgent needs you today, watching is what you are waiting on, and skip is counted but never opened. It is read-only, so nothing is touched.
- 04 Reply when you want to
Ask for a draft and your drafter writes two options in your voice. Copy the one you like into Gmail or Outlook and send it. Nothing is ever sent for you.
- 05 Optional: make it sound like you
Run build-user-context and build-inbox-voice once first. The triage then knows your business and the drafts read in your voice, not a chatbot's.
Other things you can ask.
Morning brief
Triage my inbox.
A sorted brief: what needs you today, what is waiting, and what to skip.
Your own window
Triage my inbox from the last 24 hours.
The same brief, scoped to the window you name.
Since a time
Triage my inbox since yesterday 5pm.
Working-hours aware, good for first thing after a day off.
Draft a reply
Draft a reply to the Ridgeline email.
Two drafts in your voice, ready to copy and send. Nothing is sent for you.
What is stalling
What am I still waiting on?
Threads where you sent the last message and the other side has gone quiet.
None of these is a command to memorise. The skill reads plain English and activates on its own.
What it reads and what it writes.
- Reads
- The email content you paste in or connect, and nothing else in your account.
- Writes
- A sorted brief: what needs you today, what is waiting, and what to skip.
- Never
- It sends nothing and auto-sends nothing. A reply is a separate step you ask for.
Urgent
Emails that genuinely need you today: a real ask, a deadline, or a reply to something you were chasing.
Watching
Threads where you are waiting on someone, or something brewing that is not urgent yet.
Skip
Newsletters, CC noise and automated alerts. Counted, never opened, so you see the volume without reading it.
Is this AI inbox assistant safe to use?
Yes. The skill runs inside your own AI account and reads only the email content you paste in or connect. It sorts your inbox into urgent, watching and skip and sends nothing. Drafting a reply is a separate step you ask for.
Can it send emails for me, or will it auto-send replies?
No. It sorts your inbox into three piles and never sends or auto-sends anything. When you want a reply, it hands off to your drafter, and you review and send every email yourself.
Does OpenKit or Anthropic see my inbox?
No. The skill runs in your own AI account and only ever sees the email content you choose to paste or connect. OpenKit never receives your inbox, and you control how your conversations are handled in your AI assistant privacy settings.
Its two setups.
This is the daily one. The other two are one-time setups that feed it, and running them first is what makes the brief read in your language.
Set up your inbox once, then ask each morning.
Download the folder, switch it on in Claude or ChatGPT, and hand it the morning’s mail. It sorts what came in overnight and sends nothing.
When the work is bigger than a skill, start with an audit.