Guys, WorkBuddy handles fragmented files; I turned repo notification archiving into a pipeline

Today I stumbled upon a cloudsight-ai repo. Under the main branch, there's a pile of md files with filenames like 2026-08-2523:31:17.md, looking like system logs dumped every few seconds. My first reaction was annoyance. Repos hate this kind of clutter—notifications, handovers, exceptions, reconciliation screenshots—dozens or hundreds a day that people can't possibly keep up with. I've been trying ChatExcel these past few days, used TraeWork for a week, and just touched Dataiku yesterday, but to actually turn these fragmented files into usable tables, I'm still running it through WorkBuddy.

My view is straightforward: small-to-medium repos shouldn't jump straight into building autonomous AI platforms. Get the basics stable first—file organization, summarization, and permissions. I've used WorkBuddy for four weeks, and what really saves time is its ability to bind context with permissions.

How I Set Up This Configuration

Open WorkBuddy, click 'New Project' in the left project bar, name it "Repo Notification Archive," and enter the project homepage after creation. You'll see a large file upload area; drag in recent md files, Excel sheets, email exports, and WeChat screenshots from the last day. Beginners, don't rush to batch process; start with about ten files to see if it can read them.

Then click 'Automation' at the top, go to the rules page, and click 'New Rule'. Select 'New File Enters Project' as the trigger condition. There's a term here called 'Folder Sync', which simply means binding a folder on your computer to a WorkBuddy project. From then on, any file dropped into that folder automatically appears in the project. After binding, select 'Generate Summary Table' as the action. In the field template, fill in Date, Time, Source, Team, Person in Charge, Status, and Remarks. Parse date/time from the filename, check the beginning of the file content for the source, and look for Group A / Group B in screenshots or emails for the team. The initial setup takes about twenty minutes; afterwards, new files entering will automatically populate the table.

I only focus on three key parameters: filenames must have timestamps, emails must CC a fixed public mailbox, and screenshots must include the team name. Without these three, AI will guess, but too much guessing leads to chaos.

Don't Open Permissions Too Wide

WorkBuddy has permission distribution—defining who can view, edit, or only comment. After creating the project, click 'Members' in the top right corner. Keep me as admin, set clerks as editors, and team leaders as read-only. Then click 'AI Permissions' and change the default from 'Auto Execute' to 'Manual Confirmation'. This must be changed. I learned the hard way before; wanting speed, I opened full permissions, and it mixed customer complaints and supplier reconciliations into the same table, nearly covering up an exception regarding missed payments. Later, I set a strict rule: AI can only generate suggestion tables, cannot directly move original files, and cannot directly send notifications. Wait for the clerk to confirm, then click Apply.

Daily maintenance is simple too. Every morning before work starts, I only check the exception queue on the rules page, which lists items where the team wasn't identified, dates weren't parsed, or files were duplicated. One click allows manual field completion. Export the archive table every Friday afternoon for finance and reconciliation staff. Delete expired rules once at the end of each month, such as old team names, otherwise it applies old templates to new files.

This setup isn't necessarily suitable for everyone. I have over thirty people here, lots of files but fixed processes, so it works. If you're only handling one or two spreadsheets, TraeWork or Excel might suffice. But when fragmented files pile up in a warehouse, WorkBuddy's value lies in pulling people out of the hunt for files. Don't treat it as an all-powerful AI; use it to do one small thing first—archive the last week's notifications into a single table. Once that runs smoothly, then talk about scheduling, reconciliation, and handovers.