All use cases

Engineering

Overnight agents finish in a review queue, not a write

Morning cleanup means the overnight agent crossed your irreversible line.

Share

What hurt

Wyzern (@0xWyzern) argues the safest finish state for an overnight agent is a review queue, not a changed record. Let research and reconciliation run; stop before submission. When the next step would change a customer record, hand control back. By morning: evidence collected, conflict explained, final action prepared — irreversible part untouched.

What to install or create

Define the irreversible boundary in the bot profile. Overnight: search, compare, resolve routine discrepancies, draft the consequential write. Stop before submit. Autonomy = knowing where clicking must end.

Bots

  • Overnight reconciler

    Research + draft only. Stops before system-of-record write.

Plugins

  • None on day one. Follow the first-run steps.

Files

  • Morning review queue (diffs, conflicts, proposed write).

Routines

  • Overnight: gather evidence + draft. Morning: you approve the write.

The prompt

This is a tip, not a pasteable system prompt. The author did not publish one. Follow the first-run steps.

First run

  1. 1Read https://x.com/0xWyzern/status/2097808590329250027.
  2. 2In the overnight bot profile, name the irreversible actions (submit, charge, send, delete).
  3. 3Allow research, compare, and draft through the night.
  4. 4Force stop before any system-of-record write; park output in a review queue.
  5. 5Morning: you approve or discard. Measure autonomy by knowing where clicking ends.

Approval fence

Any write that changes a customer/system-of-record row needs your morning yes.

What to tweak

Steal the finish state: review queue by morning, never a silent write to the system of record.

Source

Wyzern @0xWyzern · Original post

Not affiliated with xAI. We did not write this prompt. Third-party template terms still apply.

Browse templates

Browse team blueprints

Pro pack

Want the full setup, not just the idea?

Pro packs give you the job text, prompts, schedules and pitfalls for popular use-cases.

See Pro