Command
Use my supplied host, saved baseline, alert criteria, and days of 7, 30, or 90. Verify the project using list_projects. Read get_brand_insights({host,days}) and get_overview({host,days}); inspect concerning phrases with search_chats({host,query,days,limit:20}) using terms I supply. Compare only returned sentiment or perception measures with compatible baseline fields. If the data satisfies my criteria, draft an alert containing evidence, scope, uncertainty, and a recommended review owner. If scores or history are absent, mark the trigger unevaluable. Do not invent daily monitoring. Recurring checks and Slack or other message delivery require my independent scheduler and messaging connection.What you need
- Provide a visible host, supported window, saved baseline, and explicit alert criteria.
- Recurring execution and message delivery require separate scheduling and messaging connections.
Validate whether a trigger is measurable
Check that the returned perception measures and your baseline support the alert criteria before evaluating a change. Missing scores, incompatible definitions, or absent history remain explicit, preventing a qualitative concern from being presented as a verified numeric sentiment drop.
Give reviewers an evidence-rich alert draft
Prepare a concise message with the observed issue, sample references, limitations, and a suggested review owner. The draft supports a human decision about escalation; an independently configured scheduler and messaging system are needed to run checks repeatedly and deliver notifications.