← Back to library

Run an overnight research team with handoff notes

flash (@flasssh_ai) reworked a team layout into overnight Grok Bots: a literature reader, a reproduction lead for failed experiments, a data reviewer, a signal scout and an editor who writes a morning brief. Rules include one shared brief, one owner per stage, sourced claims and approval for sending, merging or spending.

  • OutcomeSplit research work into five agent roles that read, reproduce, reanalyze, scout and write, with one shared brief and human approval for anything outward-facing.
  • SetupAn afternoon+
  • LevelAdvanced
  • Creditflash (@flasssh_ai)
View original
Works with
Grok Bot

Published Oct 23, 2026Updated Oct 23, 2026AI dry run on sample input: passed Oct 23, 2026Source verified Oct 24, 2026

When should you use it?

  • You run experiments or follow a fast-moving literature
  • Failures pile up overnight
  • You want a morning brief with sources

Not for: One-off questions; Work where agents would need to publish or spend on their own

What do you need, and what do you get?

You need

  • Shared research brief
  • Paper feeds
  • Experiment logs
  • Data and code repo

You get

  • Morning brief separating facts from assumptions
  • Patch proposals with results
  • Handoff notes per stage

How does it work, step by step?

  1. Write one briefPut the goal, allowed sources, expected output and due time in one file every role reads.
  2. Assign stage ownersGive each role its own stage so two agents never edit the same work.
  3. Require receiptsEach claim must carry where it came from, when, and how sure the role is.
  4. Hand off in writingEach role writes down finished work, failures, and the next role's to-do.
  5. Approve in the morningAnything that sends, posts, merges, deletes or costs money waits for you.

The prompt

Our own prompt, inspired by flash (@flasssh_ai) and written to work in Grok Bot. Values in lime are placeholders.

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

Prompt3 variables
You are the {role} on an overnight research team. Shared brief: {brief}.
Your stage only: {stage_scope}. Do not edit work owned by other roles.
For every claim give source, date and confidence (high, medium, low).
Finish with a handoff note covering completed items, failures, and the to-do for the next role. Do not send, publish, merge, delete or spend.

Inspired by flash (@flasssh_ai): this is pure f*cking treasure. The prompt above is written by playbooks.bot in our own words; nothing from the original is quoted.

What can go wrong?

  • Roles overlapping on the same files cause overwrites.
  • Briefs that blend facts and guesses mislead; separate them.

Our take

The roles matter less than the rules underneath them: one shared brief, one owner per stage, written handoffs, and every claim carrying a source and confidence. Those rules make a multi-agent setup debuggable. Keep the approval boundary exactly where the post puts it. Start with two roles and a single shared brief, check morning output for a week, then add reproduction and analysis roles once handoff notes read cleanly. The written handoff is the piece that makes failures traceable: each role notes the work done, anything that broke, and the open items for whoever runs next. Keep stage ownership strict, since roles overlapping on the same files cause overwrites. Separate facts from guesses in every brief, with a source and confidence on each claim. Anything that sends, posts, merges, deletes or spends waits for your morning approval. A one-file brief that every role reads keeps the goal, allowed sources and due time consistent, and makes a bad morning brief easy to trace back to its cause.

Where the time goes back: You wake up to triaged papers, a first look at failures and a sourced brief instead of starting cold.

FAQ

Do I need five agents?
Start with a reader and an editor; add roles as the work demands.
Can it merge fixes overnight?
No. It can prepare a patch and results; merging waits for you.

Source and credit

Inspired by flash (@flasssh_ai). We describe the idea in our own words, link to the original, and wrote the prompt ourselves; nothing is quoted.

Inspired by
this is pure f*cking treasure
Creator
flash (@flasssh_ai)
Published on
X
Original date
Oct 4, 2026
License
Not openly licensed (the poster keeps all rights), so we quote nothing and link to the original
Originally for
Grok Bot
AI dry run
Oct 23, 2026: passed. Dry-run tested by our AI test agent on sample input. PASS: Gates held on desk check; no live sends, bookings or purchases.
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
Oct 24, 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

Hand off one offline task before you sign out

Before you go offline, get up to three useful task options with a concrete done artifact named for each, pick one, and wake up to a review-ready deliverable the agent prepared after your choice.

Credit: Scotty Beam (@ScottyBeamIO)

Minutes
Operations & AdminPrompt✓ AI dry run

Require the finished work plus the evidence tape

Make every agent delivery include the artifact, recordings or screenshots, linked sources, guesses called out, and skips with reasons.

Credit: Carnage (@0xCarnagee)

Minutes
Operations & AdminWorkflow✓ AI dry run

Share one to-do list between your team and your agents

Run a normal task list where people and AI agents both take tasks, update progress, and send finished work back for review.

Credit: Pete Callaghan (@petecallaghans)

An afternoon+
Operations & AdminRoutine✓ AI dry run

Trigger useful work the moment a matching email arrives

Define filters on sender, recipient or subject so an assistant summarizes, drafts, extracts or researches as soon as matching mail arrives, with results saved for review.

Credit: XFreeze (@XFreeze)

Minutes
Operations & AdminWorkflow✓ AI dry run

Get a weekday brief of only the Google Chat you missed

Start each weekday with one short summary of the Google Chat messages you have not read yet, grouped by space with a link to each, while your unread badges stay in place.

Credit: Mete Atamel

Under an hour