← Back to library

Make key agent routines leave a receipt that gets checked

Michael Fenech (@Michael_Fenech_) shared a setup for Grok Bot routines that proves the work got done: every important routine writes a receipt to a shared dashboard, and a different agent checks for missing or unexpected ones. We explain how to set it up and show his prompt word for word with credit.

  • OutcomePick the scheduled agent jobs you cannot afford to miss, have each one record a receipt with its run time, outcome and what it produced, let a separate checker compare expected runs against receipts on a schedule, and get alerted only for missing receipts, failures or odd results, after testing all four cases with harmless examples.
  • SetupUnder an hour
  • LevelIntermediate
  • CreditMichael Fenech (@Michael_Fenech_)
View original
Works with
Grok BotClaude CoworkChatGPT agentn8n

Published Nov 3, 2026Updated Nov 3, 2026AI dry run on sample input: passed Nov 3, 2026Source verified Nov 3, 2026

When should you use it?

  • Routines such as invoicing, payroll prep or client reports that would hurt if they silently skipped
  • You find out a scheduled job did not run only when someone complains
  • Several agents run scheduled work and you want one place to see what happened

Not for: Low-stakes routines where a missed day does not matter; Setups without a shared sheet, database or page the agents can all write to; Replacing proper monitoring for production software systems

What do you need, and what do you get?

You need

  • The list of routines that matter most, with schedules and time zone
  • What a normal result looks like for each and how long it should take
  • A shared dashboard, sheet or table the relevant agents can update

You get

  • A receipt step added to each chosen routine, recording owner, scheduled and finished time, outcome and output
  • Receipt deadlines per routine
  • A separate checker on its own schedule that flags missing receipts, failures and unexpected results
  • Test results for a normal run, a run with nothing to do, a failure and a missing receipt

How does it work, step by step?

  1. Choose the routinesPick only the handful of jobs whose failure would cost you money, clients or a deadline.
  2. Hand over the promptGive Fenech's prompt to your lead agent and answer its questions about schedules and expected results.
  3. Check the receiptsLook at the dashboard after the next runs and make sure each entry shows real output, not just a finished step.
  4. Run the four testsHave it simulate a normal run, an empty run, a failure and a missing receipt, and confirm the alerts.
  5. Ask about shared outagesFind out whether one outage could stop both the routines and the checker, and add an outside backup if so.

The prompt

The original prompt by Michael Fenech (@Michael_Fenech_), word for word, with a link to where it was published. Below it is our adapted version with fill-in blanks, if you want it. Works in Grok Bot, Claude Cowork, ChatGPT agent and others. Values in lime are placeholders.

Original prompt by Michael Fenech (@Michael_Fenech_)

Word for wordby Michael Fenech (@Michael_Fenech_)
Help me add receipts and a separate checker to my most important routines.

1 - Ask me which routines matter most. For each, confirm its schedule, time zone, expected result and how long it can reasonably take. Use that to set a deadline for its receipt.

2 - Create a shared dashboard that the relevant Bots can access and update. Add a final step to each selected routine that records its name, owning Bot, scheduled run time, completion time, outcome and a count or link showing what it produced. If it encounters a failure, record the reason where possible. Preserve the routine’s existing job.

3 - Treat ‘nothing needed doing’ as a valid outcome when the checks support it. Don’t mark a run successful just because it reached the final step.

4 - Have a separate checker owned by Chief of Staff compare expected runs with their receipts. Agree a checking frequency that suits the deadlines. Flag missing receipts, recorded failures and unexpected results. A missing receipt means completion is unverified, not proof that nothing ran.

5 - Only alert me when something needs attention. Include the routine, the issue, the evidence and the recommended next step. Don’t repeatedly alert me about the same unresolved issue or automatically rerun the work.

6 - Test a normal receipt, a legitimate zero-result receipt, a failure and a missing receipt using harmless examples. Explain any gaps, including whether the same outage could stop both the routines and their checker.

Exactly as published by Michael Fenech (@Michael_Fenech_) in X post: receipts and a separate checker for Grok Bot routines you can't afford to miss.

Our adapted version (optional)

Fill in the blanks (optional; the copied prompt uses your values)

Adapted prompt4 variables
Set up proof-of-run tracking for these routines: {routine_list}. For each, confirm the schedule in {time_zone}, the normal result and the latest acceptable finish time.

Create a shared log at {dashboard}. Add a last step to each routine that writes one row: routine, owner, scheduled time, finish time, result, and a count or link to what it produced, plus the error if it failed. Do not change what the routines already do.
A run that correctly found nothing to do is a normal result; finishing the steps alone is not.
Assign a different agent to compare the schedule with the log every {check_interval}. Alert me once per problem with the evidence and a suggested next step; never rerun jobs yourself.
Before going live, test a normal run, an empty run, a failure and a missing row, and tell me what a single outage could hide.

What can go wrong?

  • A receipt that only says done proves little. Require a count or link to what was produced.
  • Checker alerts that repeat every hour get ignored. Keep the rule about one alert per unresolved issue.
  • If the checker runs on the same platform, a full outage can silence both. Consider a simple external check for the most critical job.
  • Automatic reruns can duplicate payments or messages. Leave rerunning to you.

Our take

Scheduled agents fail quietly. A job that never started leaves no error, and one that finished with the wrong result can look exactly like success. Fenech's setup handles both with two simple ideas from operations work: every important run leaves a receipt showing what it produced, and someone other than the worker checks the receipts. We like the detail in his rules. An empty result counts as valid when the checks support it, reaching the last step is not proof of success, and a missing receipt means unverified rather than failed. That last distinction keeps the alerts honest. Our main caution is the one his prompt already raises: if the checker lives on the same platform as the routines, one outage can silence both, so the most critical job may deserve an outside check as well. In our dry run with a fictional agency's three routines, the setup added receipt steps and deadlines, accepted a Saturday invoice run with zero invoices as normal, flagged a payroll export that failed on a missing file, and raised one alert when the Monday client report left no receipt by its 10:00 deadline.

Where the time goes back: You stop checking every scheduled job by hand, and you find out about a skipped or failed run when it happens rather than when a client or colleague notices.

FAQ

Why a separate checker instead of letting each routine report itself?
A routine that never starts cannot report anything. A different agent comparing the schedule with the receipts notices the silence.
Is no output always a problem?
Not necessarily. The prompt treats a run with nothing to do as a valid result, as long as the checks back it up, so quiet days do not trigger alerts.
Can I build this outside Grok Bot?
Yes. The same pattern works with Claude Cowork scheduled tasks, ChatGPT agent tasks or n8n, as long as runs can write to a shared sheet or table.

Source and credit

The original prompt on this page is by Michael Fenech (@Michael_Fenech_), shown word for word with a link to where it was published. The summary, notes and adapted version are ours.

Original
X post: receipts and a separate checker for Grok Bot routines you can't afford to miss
Creator
Michael Fenech (@Michael_Fenech_)
Published on
X
Original date
Nov 2, 2026
License
Not openly licensed; the original prompt is shown word for word with credit and a link to the original
Originally for
Grok Bot
AI dry run
Nov 3, 2026: passed. Dry-run tested by our AI test agent on sample input. All four test cases behaved as specified, single alerts, no reruns, shared-outage gap reported.
AI dry run on sample data; not run with live connected accounts. The playbooks.bot test agent filled the prompt with synthetic inputs, produced the output and checked it against the page's stated outputs and guardrails.
Verified
Nov 3, 2026: source link resolves to the original, the credited creator published it, our excerpt or prompt matches it, and the license allows our use.

Are you the creator? Claim or correct this page.

More in Operations & Admin

All 203 Operations & Admin playbooks →

Related playbooks

Operations & AdminPrompt✓ AI dry run

Give your agents office hours instead of constant pings

Have your lead agent list every agent and routine that messages you, sort each message type into urgent, decision or FYI, batch decisions into two daily briefings with a recommendation each, collect routine updates into a single evening summary, and set clear rules for the rare cases that may interrupt you.

Credit: Michael Fenech (@Michael_Fenech_)

Under an hour
Operations & AdminPrompt✓ AI dry run

Audit how your agents pick tools before buying more usage

Have your lead agent review the recurring jobs it can see, note which apps, scripts, browser sessions or full computer control each one relies on, flag jobs that take a heavier route than needed or rely on remembered facts, check permissions and fallbacks, and then propose three side-by-side experiments before anything is changed.

Credit: Michael Fenech (@Michael_Fenech_)

Under an hour
Operations & AdminPrompt✓ AI dry run

Audit your agent routines for steps you can cut

Have your lead agent study one recurring routine, write down what a good result looks like, propose a shorter path to it, test old and new side by side on the same inputs, and switch only after you approve, with the old version kept for restore.

Credit: Michael Fenech (@Michael_Fenech_)

Under an hour
Operations & AdminPrompt✓ AI dry run

Check your agent can undo a bulk change before it runs

Before an agent edits hundreds of files, rows or records, get a list of risky jobs, how each would be restored, a tested undo on a copy, and a change log, with nothing live touched until you approve.

Credit: Michael Fenech (@Michael_Fenech_)

Under an hour
Operations & AdminPrompt✓ AI dry run

Make your agents read websites the cheapest reliable way

Give your lead agent a standing rule for pulling information from websites: use a connector first, then a plain script, then a browser page read by a script, and only fall back to screenshots and clicking when nothing else works, with result checks, clear failure reports and saved scripts for repeat jobs.

Credit: Michael Fenech (@Michael_Fenech_)

Under an hour