Google Sheets Row Agent

A sheet that three people and two integrations write into is fine at fifty rows and unusable at a thousand: dates in four formats, numbers no dialler will accept, and status values that nearly match what your filters look for. This agent takes each new row, repairs what has one obvious reading, in place and in that row only, and puts everything else into Slack as a question with a row number attached. It appends nothing, deletes nothing and reorders nothing, so the sheet your colleagues are working in stays where they left it.

What people use it for

• Repaired in place, never appended - The row is already in the sheet by the time this runs, so the work is repair and not entry. Nothing is duplicated, nothing shifts down, and a row number quoted in a thread still points at the same record.
• An ambiguous date stays ambiguous - 07/08/2026 in a sheet that also holds 8 July 2026 is flagged with both readings instead of guessed. A renewal firing a month early is a customer conversation nobody wanted to have.
• The header row is the contract - Columns get inserted, renamed and reordered by whoever needs them that week. Every write is resolved through the header on the day it happens, so a new column shifts nothing into the wrong cell.
• Broken exports stay visible - An email address landing in the phone column is reported as swapped rather than silently corrected. Tidying that away each month is exactly how a broken upstream integration survives a year.

Before you fork

Will it overwrite something a colleague typed deliberately?

It writes into the row that just arrived and only into cells whose value has a single possible reading, which rules out anything resembling a judgement call. Every edit lands in the spreadsheet's version history under the connected account, so you can see each one and undo any of them. If a particular column should never be touched, say so in the rules and it will only ever flag that one.

What about the rows already sitting in there?

It works on new rows as they arrive, so a sheet holding two thousand messy rows stays exactly as messy. That is deliberate, because a sweep across historic data is a much larger decision than repairing one inbound row. Most teams switch it on now and clean the history separately, once.

What do I need connected?

A Google Sheets credential with edit access on that one spreadsheet, the tab rows arrive on, and a Slack channel. Then you describe your columns and the formats you want, in your own words. Those formats do not have to match anything the sheet does today, since the rules are what it normalizes towards.