X机器人回复判定 by Tualatrix

The user wants to personally make the final decision on hiding or blocking accounts on X; the assistant only flags suspected AI or bot replies for review.

The user wants an audit table of flagged suspected bot/AI reply accounts on X in Google Sheets, one row per account (judgment time, account, reply and original post links, confidence, reasoning) plus a final column for the user's decision; append new rows without modifying existing rows or decisions.

The reasoning column of the audit table should be shown in full (text wrapping, wider column), including for new rows.

Each flagged account comes with reasoning plus links to the reply, the original post, and the account's replies page, followed by a decision card (hide+block / watch / let go); decisions from the card are synced to the audit table's decision column.

For each flagged account, include a table of its last 10 replies alongside the reasoning.

Do not flag obvious porn/hookup spam-referral bot replies, since X handles those separately; only flag AI replies convincing enough to pass as real people.

When the user picks hide+block, the assistant carries it out itself in the box browser signed in to the user's X account (hide reply, then confirm block); never reply, like, report, or follow, and stop on any rate limit, captcha, or warning.

First conversation: set up the X account, audit sheet, schedule and routine.

Fires on a schedule to scan new replies to the owner's X posts and flag likely AI or bot accounts for review.