Get a reputation report every morning
A scheduled task that sweeps social for mentions of you overnight, reads the sentiment, catches anything false, drafts the response — and emails it before you open the app.
Somebody has to check what was said about the client overnight. It is the same search every day, it takes half an hour, and the day it gets skipped is the day something needed answering.
A scheduled task is a prompt on a timer, and each run starts a fresh chat with the result in it. So the morning report is not a report feature — it is this question, asked at 08:00, by itself.
Write the prompt first
Get it right by hand in an ordinary chat before you schedule it — the same sweep asked live is knowing what's happening before you go live. Nobody is there to answer a follow-up at eight in the morning, so the instructions have to carry the whole job: the sources, the order, and what to do when there is nothing to say.
Every morning, find every public mention of Northsole from the last 24 hours on Instagram, X, TikTok and Reddit.
Report in this order:
1. One line: how many mentions, and whether that's normal, up or down.
2. The mentions worth reading, grouped by platform — who posted, how big their account is, what they said, and a link.
3. Sentiment for each: positive, neutral or negative, and one line on why.
4. Anything factually wrong about us. Quote it, give the correct fact, and say how visible it already is.
5. What to do: what to answer, what to ignore, what to escalate. Draft the replies for anything you say we should answer.
If there's nothing worth reporting, say so in one line rather than padding it out.Run it once. If the shape of the answer is wrong, fix the prompt — not the schedule.
Then set the schedule
Prompt
Scheduled → New schedule, or the "Create schedule" item on the menu of the message you just sent, which opens the form with that message already in it.
Name it something you will recognise in an inbox — Northsole morning reputation — and paste
the instructions. The instructions box is the chat composer, so @ a project to load the
client's knowledge into every run and / to call a skill by name. What you cannot do here is
attach a file: a schedule is text only.
Under the box, pick the model and narrow Tools if you want this run reaching less than everything you have.
Delivery
Repeats: Daily, At: 07:30. The time is read in your browser's timezone as you write it, and it keeps firing at that wall-clock time in that zone afterwards — it does not follow you if you travel.
Turn Email delivery on and add the recipients: people in the workspace, or any address you type, which is how the client gets it without getting a login. The subject line is the schedule's name.
Visibility defaults to Private — only you see the schedule and the chats its runs produce. Set it to Workspace if the account team should be able to read the morning's run without you forwarding it.
Review
The last step reads everything back and previews the next three runs with the timezone named out loud. Check those three dates before you save; this is the moment a "monthly on the 31st" mistake is cheap.
Creating the schedule is the authorization to spend. A run cannot pause to confirm a cost, so it does not ask — it goes ahead. Each social lookup is a paid step at a flat price, and a daily sweep across four platforms is many of them, every day, forever. Spending is still bounded by what the workspace can actually pay: a step the remaining balance cannot cover is refused and the run reports a failure rather than overdrawing. See credits and spending.
Narrow it if the bill is the problem. Two platforms instead of four, a floor on account size, and a rule about what to skip will cut the lookups without costing you the mentions that matter.
Search Instagram and X only, not TikTok or Reddit. Skip accounts under 1,000 followers unless the mention is negative or says something factually wrong about us.Read it, then keep going
The email lands, and so does everything else. The Activity tab on Scheduled collects every run with its status — Success, Failed or Running — and you can filter the feed to this one schedule. Opening a run shows its output, including anything it produced as a document, web page or PDF.
Each run also has its own chat, titled with the schedule's name and the time it ran, and you can carry on talking in it. That is where the report becomes work:
Take mention 3 and draft three replies in our voice — one short and warm, one that corrects the fact without sounding defensive, one that moves it to DMs. Then write the internal note for the client.These conversations deliberately stay out of chat history, so a daily schedule does not bury everything else. Activity is the way back in.
What will go wrong eventually
A run that fails is marked in the feed, and after three consecutive failures the schedule pauses itself and says so. Fix whatever broke and switch it back on — a connector that needs signing in again is the usual culprit.
A schedule runs as whoever created it: it sees what that person can see, uses their connectors, and spends against their limits. If they leave the workspace it is paused rather than left firing runs that can only fail. Creating one needs a role that can create content, which members and admins have by default and viewers do not.
Watching something that isn't social
This recipe is social only, on purpose. Endless reads Instagram, X, TikTok, YouTube, Reddit, LinkedIn and Facebook, and there is no built-in access to reviews, maps listings or app stores. To watch a source outside that list, it has to reach Endless through a custom MCP server you connect yourself — adding one needs permission to create content, and once it is there a schedule can use it like any other tool.