WorkBuddy keeps erroring when merging tables and screenshots; anyone know what setting is wrong?

I came across a roundup of AI tools for 2026 today. It said Manus can break down steps itself based on vague goals, and Gamma can generate briefs just from a title. My first reaction was a bit of a 'DNA moment'—it really resonated. Recently, I've been tortured by scattered data. A client threw CSV daily reports, WeChat screenshots, and an Excel file with unmerged headers at me, wanting me to organize them into a material table ready for direct upload to the official account backend. I wanted to use WorkBuddy to auto-merge them and flag duplicates and null values along the way.

I've been using WorkBuddy for almost a month now. Writing weekly reports and splitting documents usually goes smoothly, but this time I couldn't figure it out. I dragged all three files into the upload box and entered something like "Merge by date, keep source, put anomalies at the end." After spinning for a while, it warned that some fields couldn't be parsed and generated a placeholder file. When I downloaded it, the headers were there, but the content was almost empty, and the date column was garbled.

Later, I tried separating them. Uploading the CSV alone worked; Excel was recognized fine too. But putting in the screenshot resulted in an image-table recognition failure. I changed the dates to text format, but got the same result. Is mixed uploading not supported? Or does image-to-table conversion need a specific mode enabled?

Maybe I made the request too complex. I wanted it to handle merging and cleaning in one go, and I stuffed the source annotation into the same sentence. Has any veteran encountered this? Should this kind of task be thrown at it all at once, or should I break it into task cards, specifying fields, date formats, and anomaly rules separately? What scares me most is if it doesn't throw an error but silently corrupts client data. Are there any small tricks to verify WorkBuddy's merge results?