When should you use it?
- Commitments get buried in threads
- You use several communication tools
- You want a daily prioritized list
Not for: Team task assignment; Workspaces whose admins forbid third-party connectors
What do you need, and what do you get?
You need
- A paid Claude plan
- A connector service linking your apps
- A to-do list location
You get
- A merged daily to-do list
- Urgency highlights
- Reminders for overdue items
How does it work, step by step?
- Connect toolsLink email, calendar and Slack through a connector service.
- Teach the skillExplain in plain English what counts as an action item for you.
- Run each morningScan recent messages and notes for commitments and requests.
- Merge and rankAdd new items to your list and flag the urgent ones.
- Let it rememberHave it keep notes on what you mark as done or irrelevant.
The prompt
Our own prompt, inspired by Claire Morrison and written to work in Claude Cowork and Claude Code. Values in lime are placeholders.
Each weekday at 8am, scan my Slack messages, email and notes from the last 24 hours for things I agreed to do or was asked to do. Add new items to [to-do list] with the source link and any due date. Highlight the three most urgent. Do not reply to anyone or mark anything as read.
Inspired by Claire Morrison: Create your personal agent with Claude. The prompt above is written by playbooks.bot in our own words; nothing from the original is quoted.
What can go wrong?
- Connector services hold broad access; check what each one can do.
- Vague requests in chat can be over-counted as tasks; review the list.
Our take
Claire Morrison and her co-author's daily sweep is a strong first job for a personal work agent because it mostly reads rather than acts. Pulling action items from Slack, email and notes into one ranked list removes the tax of hunting through tabs each morning. Have it link each item back to the source message so you can check context in one click, and merge duplicates when the same request appears in two channels. Review connector permissions closely and keep send rights off at first; a triage agent does not need to reply. After a week, tell it which items were noise and which mattered, because that feedback is what turns a dump of bullets into a useful list. Skills taught in plain English are fine to start, then tighten wording where it misfiles things. Keep calendar creates and message sends behind your approval until the ranking has proven accurate for several weeks.
Where the time goes back: Hunting for what you promised across tools takes time every morning; a merged list is ready before you start.
FAQ
Do I need to code?
What else does their agent do?
Source and credit
Inspired by Claire Morrison. We describe the idea in our own words, link to the original, and wrote the prompt ourselves; nothing is quoted.
- Inspired by
- Create your personal agent with Claude
- Creator
- Claire Morrison
- Published on
- technicallycurious.substack.com (Substack)
- Original date
- Feb 2, 2026
- License
- Not openly licensed (the poster keeps all rights), so we quote nothing and link to the original
- Originally for
- Claude Cowork
- AI dry run
- Oct 23, 2026: passed. Dry-run tested by our AI test agent on sample input. PASS: Gates held on desk check; no live sends or purchases.
AI dry run on sample data; not run with live connected accounts. The playbooks.bot test agent filled the prompt with synthetic inputs, produced the output and checked it against the page's stated outputs and guardrails. - Verified
- Oct 23, 2026: source link resolves to the original, the credited creator published it, our excerpt or prompt matches it, and the license allows our use.
Are you the creator? Claim or correct this page.