POHA workflow
A Claude Desktop template where nine scheduled tasks build a morning brief from email, WhatsApp and calendar.
- 122 ★ on the repository
- Last commit
Broad access
Why this level
- The agent reads personal messages, email, finances and health data if the matching connectors are attached.
- Tasks run on a schedule without your involvement and send emails to your address.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add workflows/pohaDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent reads personal messages, email, finances and health data if the matching connectors are attached.
- Tasks run on a schedule without your involvement and send emails to your address.
This is third-party code. Review the repository files before installing.
What it does
POHA, the Personal Overnight Helper Agent, is a set of ready prompts and memory for Claude Desktop in Cowork mode. Nine scheduled tasks read inbound items from WhatsApp, Gmail, the calendar and task lists. The 5:00 morning brief arrives by email with one-tap links for each action item. The rest of the schedule covers an evening reflection, a Sunday weekly review, turning agreements into calendar events, a Friday agenda, a weekend event search, and a monthly financial sweep and health memo. Five commands run on demand: `/draft`, `/reply`, `/wildcard`, `/gsd` and `/roast`. Memory lives in plain markdown files with size budgets, and the `tools/memory_maint.py` script enforces owners and max age. A five-tier system, from family to other messages, sets attention priority.
Who it is for. Claude Pro or Max users with a busy life of family, work and projects who lose conversation threads and want a brief before the alarm.
Good fit when
- A "let us meet Saturday" agreement from a chat should become a calendar event.
- Several groups and email threads went unanswered and a morning brief is needed.
- A Friday agenda is needed from tasks and open commitments.
- A thread reply is wanted in three options with a recommendation.
Not a fit when
- The model must not read personal messages and email.
- You want a chatbot to talk to rather than a background agent.
- Spending about three hours on the initial memory setup is not acceptable.
Example request
Read CLAUDE.md and walk me through filling in the memory files, then set up the seven scheduled tasks from BACKLOG.md.Limitations
It needs Claude Desktop with Cowork mode and a Claude Pro or Max subscription, with Max recommended for all tasks. Some connectors cost money, such as Monarch Money for the financial sweep, while the rest are free for personal accounts. Initial memory setup takes about three hours. The agent reads messages and email by design, so open `docs/privacy.md` before connecting finance or medical data. The repository is a personal template, and the placeholders in `CLAUDE.md` are filled in by hand.
How to disable. Delete the scheduled tasks in Cowork and disconnect the connectors. The memory folder sits on your machine and is deleted like any ordinary folder.
FAQ
What does POHA need to run?
Claude Desktop with Cowork mode, a Claude Pro or Max subscription and a clone of the repository in a stable folder. Then fill in `CLAUDE.md` and connect connectors, at minimum Gmail and calendar.
Where is the data stored?
Memory lives in markdown files on your machine, and `memory/` is excluded from Git by default. Only Claude API calls and the briefing emails you send to yourself leave it.
How does the agent learn an offline action is done?
Each briefing email carries `mailto:` links next to action items. Tapping one sends you an email with a coded subject that the next briefing parses to close the item.
Can I drop tasks I do not need?
Yes. Each file in `briefings/` is a standalone prompt: edit them, remove the ones you do not want and add your own.
What do the on-demand commands do?
`/draft` writes an outbound message in your voice, `/reply` gives three replies with a recommendation, `/wildcard` offers a contrarian insight, `/gsd` builds a priority list and `/roast` roasts a group chat.
Discussion
No questions or reviews yet.
Related
WorkflowsAn Obsidian vault template for Claude Code: /daily, /weekly, /monthly and /project commands link three-year goals to today's tasks
An Obsidian vault with skills for work meetings: Gmail transcripts, participants, follow-up tasks and weekly recaps.
A personal AI chief of staff on a notes folder: daily plan, meeting prep, tasks and people context in Claude Code and Cursor.
About a hundred skills for Claude Code: meetings, research, publishing, TDD, Telegram, email and personal analytics
Ready-made Obsidian vault with Claude Code skills for projects, tasks, clients and daily reviews, started with /setup-memory.
Claude Code Toolkit (robertguss)
Claude Code Toolkit
A set of Claude Code skills and hooks: brainstorming, worker-led project leadership, a wiki, and a book pipeline.
README in short
The architecture has three layers: a scheduled briefing engine, markdown memory and on-demand commands. Three knobs change behavior: the persona in `CLAUDE.md`, the tier system and the task schedule.