← Back to library

Brief your agent on the outcome, not every step

Michael Fenech (@Michael_Fenech_) argues that many Grok Bot setups are slow and costly because owners script every step and tool. His five-rule brief starts from the result you want and leaves the route to the agent, with approval points and a final check built in. We explain how to use it and show his prompt word for word with credit.

  • OutcomeHand a new job to your lead agent by describing the finished result instead of the steps, and have it confirm what success looks like, ask only for context it cannot find itself, pick the simplest dependable approach, state what it will do alone and what needs your approval, and check the result against your goal before calling it done.
  • SetupMinutes
  • LevelStarter
  • CreditMichael Fenech (@Michael_Fenech_)
View original
Works with
Grok BotClaude CoworkChatGPT agentCustom agent

Published Nov 4, 2026Updated Nov 4, 2026AI dry run on sample input: passed Nov 4, 2026Source verified Nov 4, 2026

When should you use it?

  • You are about to give an agent a job it has not done before
  • Your instructions for agents keep growing into long step-by-step scripts
  • Usage or run times are creeping up and you suspect the agent follows a longer route than needed

Not for: Regulated or audited processes where every step must follow a fixed procedure; Jobs where you already know the one correct method and it rarely changes; Agents without access to the files, apps or connections the job needs

What do you need, and what do you get?

You need

  • A one or two sentence description of the result you want
  • Any hard limits, such as budget, deadline, or people and systems it must not contact
  • The connections and files the agent can already reach

You get

  • A definition of the finished result and how success will be judged
  • A short list of questions the agent could not answer from its own sources
  • A proposed approach with the approval points marked
  • A closing report of what was done, what could not be confirmed and whether to make it a repeatable workflow

How does it work, step by step?

  1. Write the goalDescribe the finished result in plain words, not the method, and paste it into the bracket in Fenech's prompt.
  2. Answer only real gapsReply to the agent's questions, and tell it where to look if it asks for something it can already find.
  3. Approve the planRead the proposed approach and the list of actions that need your sign-off before it starts.
  4. Review the checkWhen it reports back, compare the result with your goal and look closely at anything marked unverified.
  5. Decide on repeating itIf it suggests turning the job into a routine, agree only once the first result holds up.

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 achieve the following outcome: [DESCRIBE YOUR GOAL].

Use these five rules:

1: OUTCOME FIRST — Clarify exactly what the finished result should look like and how we'll know it's successful.

2: CONTEXT — Ask for any essential information, files, connections or preferences you need. Don't ask questions you can answer from reliable sources you already have access to.

3: CHOOSE THE PATH — Work out the simplest reliable approach. Prefer existing connections and reusable methods where appropriate. Don't add unnecessary steps, Bots, routines or tools.

4: SET BOUNDARIES — Explain what you can handle independently, what needs my approval and what you must never do. Don't take irreversible or external actions without permission.

5: VERIFY THE RESULT — Check the actual outcome against my goal. Tell me what's completed, what's unverified and whether a simpler repeatable workflow would make sense.

Before starting, give me a short summary of your proposed approach and ask for approval where required.

Focus on delivering the outcome, not blindly following a long list of instructions.

Exactly as published by Michael Fenech (@Michael_Fenech_) in X post: a five-rule outcome-first brief for new Grok Bot jobs.

Our adapted version (optional)

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

Adapted prompt3 variables
Here is the result I need: {goal}. It is for {audience} and is due {deadline}.

Before doing anything, write two or three sentences on what the finished work looks like and how we will judge it. List only the questions you cannot answer from the files and apps you already have.
Propose the shortest dependable way to get there using what is already connected, without creating new agents or schedules unless I agree.
Mark which steps you will take on your own and which need my approval; anything that sends, pays, deletes or publishes always needs it.
When finished, compare the result with the goal and tell me what is done, what you could not confirm and whether this should become a repeatable job.

What can go wrong?

  • A vague goal gives the agent room to solve a different problem. Say who the result is for and what done means.
  • Freedom to pick the route is not permission to act. Keep the rule about irreversible and external actions.
  • Agents can claim a check they did not do. Ask how it verified the result, not only whether it did.
  • Some jobs really do need a fixed procedure. Keep step lists where compliance or safety depends on them.

Our take

This is a small prompt with a big idea behind it: the person handing out the work should define the destination and the guardrails, not the turn-by-turn route. Long step-by-step instructions feel safe, but they lock an agent into whatever method you happened to imagine, even when a connected app or an existing method would do the job in fewer moves. Fenech's five rules keep the safety that scripts were supposed to provide. The agent has to define success up front, explain what it may do alone and what needs your yes, and finish by checking the actual result rather than reporting that it followed the plan. We also like the instruction not to ask questions it can answer itself, which cuts the back-and-forth that makes delegation tiring. Our caution is that a loose goal invites a loose result, so spend your effort on describing done precisely. In our dry run with a fictional bookkeeping firm asking for year-end thank-you emails, the agent defined success as one draft per active client with a correct project count, asked only about tone and inactive clients, chose the connected CRM over a new routine, and held every send for approval.

Where the time goes back: You spend a minute describing the result instead of half an hour writing a script, and the agent avoids extra steps, tools and helper agents that a step-by-step plan can force on it.

FAQ

Does this replace detailed instructions completely?
No. It moves the detail to the result, the limits and the approval points. You can still name a method when it matters, for example a required template.
Why should it save usage?
When you dictate every tool and step, the agent may follow a longer path than it would choose. Fenech's point is that a clear goal lets it use existing connections and skip steps that add nothing. Measure it on your own jobs.
Will it work outside Grok Bot?
Yes. Any agent that can plan before acting, such as Claude Cowork or ChatGPT agent, can follow the same five rules.

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: a five-rule outcome-first brief for new Grok Bot jobs
Creator
Michael Fenech (@Michael_Fenech_)
Published on
X
Original date
Nov 3, 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 4, 2026: passed. Dry-run tested by our AI test agent on sample input. Success defined up front, asked only for missing context, picked the connected CRM over new tools, held sends for approval, verification plan matched the goal.
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 4, 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

Audit how you manage your agent team, not only the agents

Have your chief-of-staff agent review the last 30 days of how you delegated, clarified, interrupted, corrected, assigned, approved and followed up across your agent team, score eight management areas with evidence, list your three biggest weaknesses, and draft a seven-day improvement plan, without changing any bot, routine or instructions until you approve.

Credit: Michael Fenech (@Michael_Fenech_)

Under an hour
Operations & AdminPrompt✓ AI dry run

Brief an agent with nine fields, not a how-to

Before you hand a job to an agent, fill in nine short fields that name the result, the sources it may use, the deadline, what it may do alone, what needs your say-so, what it must never do, how to prove the work is finished, what to do when it gets stuck, and where to put the deliverable, then add a line that tells it to pick the simplest reliable method.

Credit: Michael Fenech (@Michael_Fenech_)

Minutes
Operations & AdminPrompt✓ AI dry run

Build a dashboard page your agent refreshes on approval

Have your agent interview you about what a dashboard should track, where each number comes from and how fresh it must be, then build one page with the most important items on top, a last-checked time on every section, clear flags for missing or stale data, and a proposed refresh schedule that waits for your approval.

Credit: Michael Fenech (@Michael_Fenech_)

Under an hour
Operations & AdminPrompt✓ AI dry run

Check which of your agents should be merged or retired

Have your lead agent map every agent on your team with its jobs, routines, skills, connections and recent activity, find overlap and unused output, recommend keep, absorb, retire or hand back to you for each, and draft a safe wind-down plan covering routines, open work, saved instructions, files and sign-ins, with nothing removed until you approve.

Credit: Michael Fenech (@Michael_Fenech_)

Under an hour
Operations & AdminRoutine✓ AI dry run

Get a weekly performance report on your rental listings

Have an agent review your short-term rental pricing and bookings every week and send a short report with suggested price changes you approve.

Credit: madaxe_again on Hacker News

Under an hour