Use it again: from prompt to workflow
In this chapter you saw how a weak ask becomes a builder brief, and you saved a reusable tool card. Those are rungs of the same ladder, and they have names:
- A prompt is a request. You type it, you get an answer, and tomorrow you start from zero.
- A brief is a job spec. It defines what good looks like before any output exists.
- A workflow is a system that remembers. The brief, the rules, the examples, the checks — saved, reused, and improved every time it runs.
Prompts are disposable. Workflows compound. And when a workflow needs files and moving parts, it becomes a project — that was the "from prompt to project" lesson. Same ladder, one more rung.
But a tool card you have only used once is still a guess. The proof is the second run.
The second run
Take the tool card you built earlier in this chapter (Tracker export → weekly status update) and run it on this input. Skipped that step, or your card is blank? Borrow this minimal one:
Tool name: Tracker export → weekly status update
Rules it follows: unresolved questions go in their own list,
not as action items.
Output format: slipped items (owner, date), open questions,
follow-up email draft.
Checks before trusting: did it invent anything? miss anything?
(It is deliberately missing a rule or two. That is the point of the second run.)
This input is messier on purpose:
weds sync - sarah, dev, jamal (jamal late)
budget?? dev thinks we're over, sarah says wait for finance
new onboarding flow APPROVED - ship next sprint
someone needs to talk to legal about the data thing
jamal - hiring update friday (maybe)
also: offsite moved AGAIN, no date
This run happens off this page. Paste your card's rules and output format into your AI tool, then this input, and look hard at the result. If you want a copy on this device, type it in the notes field below or park a thought. Then continue — no sign-in required.
Check the work
Three questions, every time:
- Did it invent anything? Real notes are full of holes. If the output assigned an owner to "someone needs to talk to legal," that is fiction, not extraction.
- Did it miss anything? Is the unresolved budget question in there? The offsite with no date?
- Could you actually use it? Would you send the follow-up email it drafted, as is, with your name on it?
If any answer is wrong — good. That is not a failed run. That is the tool telling you which rule it is missing.
Use it again: from prompt to workflow
In this chapter you saw how a weak ask becomes a builder brief, and you saved a reusable tool card. Those are rungs of the same ladder, and they have names:
- A prompt is a request. You type it, you get an answer, and tomorrow you start from zero.
- A brief is a job spec. It defines what good looks like before any output exists.
- A workflow is a system that remembers. The brief, the rules, the examples, the checks — saved, reused, and improved every time it runs.
Prompts are disposable. Workflows compound. And when a workflow needs files and moving parts, it becomes a project — that was the "from prompt to project" lesson. Same ladder, one more rung.
But a tool card you have only used once is still a guess. The proof is the second run.
The second run
Take the tool card you built earlier in this chapter (Incident notes → weekly follow-up list) and run it on this input. Skipped that step, or your card is blank? Borrow this minimal one:
Tool name: Incident notes → weekly follow-up list
Rules it follows: actions whose service is not on the approved list go
in their own list, not as this-week ships.
Output format: this-week follow-ups (service, owner), open questions,
email draft.
Checks before trusting: did it invent anything? miss anything?
(It is deliberately missing a rule or two. That is the point of the second run.)
This input is messier on purpose:
monday follow-up - payments, search
error rate on payments?? still climbing
docs typo APPROVED to backlog
someone needs to pick the two ships this week
jamal - service list friday (maybe)
also: action with no service on the approved list
This run happens off this page. Paste your card's rules and output format into your AI tool, then this input, and look hard at the result. If you want a copy on this device, type it in the notes field below or park a thought. Then continue — no sign-in required.
Check the work
Three questions, every time:
- Did it invent anything? Real notes are full of holes. If the output shipped an action whose service was not on the list, that is fiction, not extraction.
- Did it miss anything? Is the climbing error rate in there? The action with no approved service?
- Could you actually use it? Would you send the follow-up email it drafted, as is, with your name on it?
If any answer is wrong — good. That is not a failed run. That is the tool telling you which rule it is missing.
Use it again: from prompt to workflow
In this chapter you saw how a weak ask becomes a builder brief, and you saved a reusable tool card. Those are rungs of the same ladder, and they have names:
- A prompt is a request. You type it, you get an answer, and tomorrow you start from zero.
- A brief is a job spec. It defines what good looks like before any output exists.
- A workflow is a system that remembers. The brief, the rules, the examples, the checks — saved, reused, and improved every time it runs.
Prompts are disposable. Workflows compound. And when a workflow needs files and moving parts, it becomes a project — that was the "from prompt to project" lesson. Same ladder, one more rung.
But a tool card you have only used once is still a guess. The proof is the second run.
The second run
Take the tool card you built earlier in this chapter (Asset list → campaign ship-list) and run it on this input. Skipped that step, or your card is blank? Borrow this minimal one:
Tool name: Asset list → campaign ship-list
Rules it follows: assets without a ship date go in their own list,
not on the lock list.
Output format: slip picks (channel, date), open questions,
lock-email draft.
Checks before trusting: did it invent anything? miss anything?
(It is deliberately missing a rule or two. That is the point of the second run.)
This input is messier on purpose:
monday lock - email, paid, landing
legal on the claim?? still pending
badge line APPROVED - keep
someone needs to pick which two assets slip
jamal - channel owner friday (maybe)
also: landing page has no ship date
This run happens off this page. Paste your card's rules and output format into your AI tool, then this input, and look hard at the result. If you want a copy on this device, type it in the notes field below or park a thought. Then continue — no sign-in required.
Check the work
Three questions, every time:
- Did it invent anything? Real notes are full of holes. If the output put a dateless asset on the lock list, that is fiction, not extraction.
- Did it miss anything? Is the pending legal claim in there? The landing page with no date?
- Could you actually use it? Would you send the follow-up email it drafted, as is, with your name on it?
If any answer is wrong — good. That is not a failed run. That is the tool telling you which rule it is missing.
Use it again: from prompt to workflow
In this chapter you saw how a weak ask becomes a builder brief, and you saved a reusable tool card. Those are rungs of the same ladder, and they have names:
- A prompt is a request. You type it, you get an answer, and tomorrow you start from zero.
- A brief is a job spec. It defines what good looks like before any output exists.
- A workflow is a system that remembers. The brief, the rules, the examples, the checks — saved, reused, and improved every time it runs.
Prompts are disposable. Workflows compound. And when a workflow needs files and moving parts, it becomes a project — that was the "from prompt to project" lesson. Same ladder, one more rung.
But a tool card you have only used once is still a guess. The proof is the second run.
The second run
Take the tool card you built earlier in this chapter (Export folder → brand-review deck) and run it on this input. Skipped that step, or your card is blank? Borrow this minimal one:
Tool name: Export folder → brand-review deck
Rules it follows: assets missing from the kit list go in their own list,
not as approved.
Output format: redo picks (asset, kit rule), open questions,
review note draft.
Checks before trusting: did it invent anything? miss anything?
(It is deliberately missing a rule or two. That is the point of the second run.)
This input is messier on purpose:
tuesday review - wordmark, banner, icon set
banner size?? export is 2x the max width
icon set APPROVED - matches the kit
someone needs to check the licensed photo folder
jamal - redo note friday (maybe)
also: new purple on the logo, not on the kit list
This run happens off this page. Paste your card's rules and output format into your AI tool, then this input, and look hard at the result. If you want a copy on this device, type it in the notes field below or park a thought. Then continue — no sign-in required.
Check the work
Three questions, every time:
- Did it invent anything? Real notes are full of holes. If the output approved an asset missing from the kit list, that is fiction, not extraction.
- Did it miss anything? Is the oversized banner in the redo list? The new purple not on the kit?
- Could you actually use it? Would you send the follow-up email it drafted, as is, with your name on it?
If any answer is wrong — good. That is not a failed run. That is the tool telling you which rule it is missing.
Use it again: from prompt to workflow
In this chapter you saw how a weak ask becomes a builder brief, and you saved a reusable tool card. Those are rungs of the same ladder, and they have names:
- A prompt is a request. You type it, you get an answer, and tomorrow you start from zero.
- A brief is a job spec. It defines what good looks like before any output exists.
- A workflow is a system that remembers. The brief, the rules, the examples, the checks — saved, reused, and improved every time it runs.
Prompts are disposable. Workflows compound. And when a workflow needs files and moving parts, it becomes a project — that was the "from prompt to project" lesson. Same ladder, one more rung.
But a tool card you have only used once is still a guess. The proof is the second run.
The second run
Take the tool card you built earlier in this chapter (Ticket export → weekly queue report) and run it on this input. Skipped that step, or your card is blank? Borrow this minimal one:
Tool name: Ticket export → weekly queue report
Rules it follows: queues without a wait-time delta go in their own list,
not as the extra-agent pick.
Output format: extra-agent pick, wait-time deltas, open questions,
huddle note draft.
Checks before trusting: did it invent anything? miss anything?
(It is deliberately missing a rule or two. That is the point of the second run.)
This input is messier on purpose:
monday huddle - floor A slammed, floor B quiet
wait times?? export missing two queues
billing queue APPROVED for the extra agent - confirm with lead
someone needs to check the overflow line
jamal - on-call swap friday (maybe)
also: new queue name in the export, no owner
This run happens off this page. Paste your card's rules and output format into your AI tool, then this input, and look hard at the result. If you want a copy on this device, type it in the notes field below or park a thought. Then continue — no sign-in required.
Check the work
Three questions, every time:
- Did it invent anything? Real notes are full of holes. If the output named a queue that was missing from the export, that is fiction, not extraction.
- Did it miss anything? Is the missing wait-time delta in there? The new queue with no owner?
- Could you actually use it? Would you send the follow-up email it drafted, as is, with your name on it?
If any answer is wrong — good. That is not a failed run. That is the tool telling you which rule it is missing.
Use it again: from prompt to workflow
In this chapter you saw how a weak ask becomes a builder brief, and you saved a reusable tool card. Those are rungs of the same ladder, and they have names:
- A prompt is a request. You type it, you get an answer, and tomorrow you start from zero.
- A brief is a job spec. It defines what good looks like before any output exists.
- A workflow is a system that remembers. The brief, the rules, the examples, the checks — saved, reused, and improved every time it runs.
Prompts are disposable. Workflows compound. And when a workflow needs files and moving parts, it becomes a project — that was the "from prompt to project" lesson. Same ladder, one more rung.
But a tool card you have only used once is still a guess. The proof is the second run.
The second run
Take the tool card you built earlier in this chapter (Claims spreadsheet → clearance note) and run it on this input. Skipped that step, or your card is blank? Borrow this minimal one:
Tool name: Claims spreadsheet → clearance note
Rules it follows: claims without a source stay in their own list,
not in the cleared lines.
Output format: cleared lines, struck lines (reason), open questions,
clearance note draft.
Checks before trusting: did it invent anything? miss anything?
(It is deliberately missing a rule or two. That is the point of the second run.)
This input is messier on purpose:
Q3 launch — "fastest in class" ?? check last year's deck
sarah wants "used by 40,000 teams" on the homepage
the G2 badge line is FINE — keep
someone said the SOC2 claim expired in March
dev thinks we can say "realtime" — legal has not signed
also: competitor comparison table, two cells still blank
This run happens off this page. Paste your card's rules and output format into your AI tool, then this input, and look hard at the result. If you want a copy on this device, type it in the notes field below or park a thought. Then continue — no sign-in required.
Check the work
Three questions, every time:
- Did it invent anything? Real notes are full of holes. If the output marked a blank comparison cell as cleared, that is fiction, not extraction.
- Did it miss anything? Is the expired SOC2 claim in the struck list? The two blank comparison cells?
- Could you actually use it? Would you send the follow-up email it drafted, as is, with your name on it?
If any answer is wrong — good. That is not a failed run. That is the tool telling you which rule it is missing.
Use it again: from prompt to workflow
In this chapter you saw how a weak ask becomes a builder brief, and you saved a reusable tool card. Those are rungs of the same ladder, and they have names:
- A prompt is a request. You type it, you get an answer, and tomorrow you start from zero.
- A brief is a job spec. It defines what good looks like before any output exists.
- A workflow is a system that remembers. The brief, the rules, the examples, the checks — saved, reused, and improved every time it runs.
Prompts are disposable. Workflows compound. And when a workflow needs files and moving parts, it becomes a project — that was the "from prompt to project" lesson. Same ladder, one more rung.
But a tool card you have only used once is still a guess. The proof is the second run.
The second run
Take the tool card you built earlier in this chapter (Cleaned export → research readout) and run it on this input. Skipped that step, or your card is blank? Borrow this minimal one:
Tool name: Cleaned export → research readout
Rules it follows: cuts without a receipt go in their own list,
not as findings to act on.
Output format: finding to act on, caveats, open questions,
readout draft.
Checks before trusting: did it invent anything? miss anything?
(It is deliberately missing a rule or two. That is the point of the second run.)
This input is messier on purpose:
wednesday readout - three cuts, one decision
row count on cut B?? dropped, no receipt
cut A APPROVED - receipt attached
someone needs to check the stale source file
jamal - method note friday (maybe)
also: segment name not on the approved cut list
This run happens off this page. Paste your card's rules and output format into your AI tool, then this input, and look hard at the result. If you want a copy on this device, type it in the notes field below or park a thought. Then continue — no sign-in required.
Check the work
Three questions, every time:
- Did it invent anything? Real notes are full of holes. If the output named a finding whose receipt was missing, that is fiction, not extraction.
- Did it miss anything? Is the stale source file in the caveats? The segment not on the cut list?
- Could you actually use it? Would you send the follow-up email it drafted, as is, with your name on it?
If any answer is wrong — good. That is not a failed run. That is the tool telling you which rule it is missing.
Use it again: from prompt to workflow
In this chapter you saw how a weak ask becomes a builder brief, and you saved a reusable tool card. Those are rungs of the same ladder, and they have names:
- A prompt is a request. You type it, you get an answer, and tomorrow you start from zero.
- A brief is a job spec. It defines what good looks like before any output exists.
- A workflow is a system that remembers. The brief, the rules, the examples, the checks — saved, reused, and improved every time it runs.
Prompts are disposable. Workflows compound. And when a workflow needs files and moving parts, it becomes a project — that was the "from prompt to project" lesson. Same ladder, one more rung.
But a tool card you have only used once is still a guess. The proof is the second run.
The second run
Take the tool card you built earlier in this chapter (Tracker export → Q3 roadmap deck) and run it on this input. Skipped that step, or your card is blank? Borrow this minimal one:
Tool name: Tracker export → Q3 roadmap deck
Rules it follows: initiatives without a launch date go in their own list,
not as on-track.
Output format: scope-cut picks (team, date), open questions,
director note draft.
Checks before trusting: did it invent anything? miss anything?
(It is deliberately missing a rule or two. That is the point of the second run.)
This input is messier on purpose:
Q3 lock - two teams, one date
staffing gap?? no new owner on the dependency
compliance review APPROVED - keep the date
someone needs to name the two scope cuts
jamal - risk register friday (maybe)
also: launch date missing on the shared tracker row
This run happens off this page. Paste your card's rules and output format into your AI tool, then this input, and look hard at the result. If you want a copy on this device, type it in the notes field below or park a thought. Then continue — no sign-in required.
Check the work
Three questions, every time:
- Did it invent anything? Real notes are full of holes. If the output marked a dateless initiative on-track, that is fiction, not extraction.
- Did it miss anything? Is the staffing gap in the open questions? The tracker row with no launch date?
- Could you actually use it? Would you send the follow-up email it drafted, as is, with your name on it?
If any answer is wrong — good. That is not a failed run. That is the tool telling you which rule it is missing.
Use it again: from prompt to workflow
In this chapter you saw how a weak ask becomes a builder brief, and you saved a reusable tool card. Those are rungs of the same ladder, and they have names:
- A prompt is a request. You type it, you get an answer, and tomorrow you start from zero.
- A brief is a job spec. It defines what good looks like before any output exists.
- A workflow is a system that remembers. The brief, the rules, the examples, the checks — saved, reused, and improved every time it runs.
Prompts are disposable. Workflows compound. And when a workflow needs files and moving parts, it becomes a project — that was the "from prompt to project" lesson. Same ladder, one more rung.
But a tool card you have only used once is still a guess. The proof is the second run.
The second run
Take the tool card you built earlier in this chapter (Interview notes → debrief packet) and run it on this input. Skipped that step, or your card is blank? Borrow this minimal one:
Tool name: Interview notes → debrief packet
Rules it follows: scores missing a competency go in their own list,
not as advance calls.
Output format: advance pick, holds (competency), open questions,
panel note draft.
Checks before trusting: did it invent anything? miss anything?
(It is deliberately missing a rule or two. That is the point of the second run.)
This input is messier on purpose:
friday panel - three candidates, one seat
score on systems?? sheet blank for candidate B
candidate A APPROVED to advance - two receipts
someone needs to probe the missing competency
jamal - follow-up friday (maybe)
also: debrief says maybe, no owner on the next step
This run happens off this page. Paste your card's rules and output format into your AI tool, then this input, and look hard at the result. If you want a copy on this device, type it in the notes field below or park a thought. Then continue — no sign-in required.
Check the work
Three questions, every time:
- Did it invent anything? Real notes are full of holes. If the output advanced a candidate with a blank required score, that is fiction, not extraction.
- Did it miss anything? Is the blank systems score in the holds? The maybe with no owner?
- Could you actually use it? Would you send the follow-up email it drafted, as is, with your name on it?
If any answer is wrong — good. That is not a failed run. That is the tool telling you which rule it is missing.
Use it again: from prompt to workflow
In this chapter you saw how a weak ask becomes a builder brief, and you saved a reusable tool card. Those are rungs of the same ladder, and they have names:
- A prompt is a request. You type it, you get an answer, and tomorrow you start from zero.
- A brief is a job spec. It defines what good looks like before any output exists.
- A workflow is a system that remembers. The brief, the rules, the examples, the checks — saved, reused, and improved every time it runs.
Prompts are disposable. Workflows compound. And when a workflow needs files and moving parts, it becomes a project — that was the "from prompt to project" lesson. Same ladder, one more rung.
But a tool card you have only used once is still a guess. The proof is the second run.
The second run
Take the tool card you built earlier in this chapter (Shift notes → handoff board) and run it on this input. Skipped that step, or your card is blank? Borrow this minimal one:
Tool name: Shift notes → handoff board
Rules it follows: steps with an open approval gate go in their own list,
not as clear.
Output format: blocked handoff, open gates, open questions,
shift note draft.
Checks before trusting: did it invent anything? miss anything?
(It is deliberately missing a rule or two. That is the point of the second run.)
This input is messier on purpose:
night shift - steps 3 through 7
gate on step 4?? still open
step 6 APPROVED - next shift named
someone needs to name the blocked handoff
jamal - exception friday (maybe)
also: customer name in the shared note, no redaction
This run happens off this page. Paste your card's rules and output format into your AI tool, then this input, and look hard at the result. If you want a copy on this device, type it in the notes field below or park a thought. Then continue — no sign-in required.
Check the work
Three questions, every time:
- Did it invent anything? Real notes are full of holes. If the output marked an open gate as clear, that is fiction, not extraction.
- Did it miss anything? Is the open gate on step 4 in there? The customer name that should be off the page?
- Could you actually use it? Would you send the follow-up email it drafted, as is, with your name on it?
If any answer is wrong — good. That is not a failed run. That is the tool telling you which rule it is missing.
Use it again: from prompt to workflow
In this chapter you saw how a weak ask becomes a builder brief, and you saved a reusable tool card. Those are rungs of the same ladder, and they have names:
- A prompt is a request. You type it, you get an answer, and tomorrow you start from zero.
- A brief is a job spec. It defines what good looks like before any output exists.
- A workflow is a system that remembers. The brief, the rules, the examples, the checks — saved, reused, and improved every time it runs.
Prompts are disposable. Workflows compound. And when a workflow needs files and moving parts, it becomes a project — that was the "from prompt to project" lesson. Same ladder, one more rung.
But a tool card you have only used once is still a guess. The proof is the second run.
The second run
Take the tool card you built earlier in this chapter (Contract clauses → privilege-review memo) and run it on this input. Skipped that step, or your card is blank? Borrow this minimal one:
Tool name: Contract clauses → privilege-review memo
Rules it follows: passages without a cite go in their own list,
not as privilege calls.
Output format: privilege calls (cite, matter), open questions,
partner note draft.
Checks before trusting: did it invent anything? miss anything?
(It is deliberately missing a rule or two. That is the point of the second run.)
This input is messier on purpose:
friday memo - matter 441, two partners
cite for §3.2?? associate thinks it's privileged, no log row
redaction of the home address APPROVED
someone needs to check the second memo that cites the same draft
jamal - privilege log update friday (maybe)
also: client name in a footer, no call yet
This run happens off this page. Paste your card's rules and output format into your AI tool, then this input, and look hard at the result. If you want a copy on this device, type it in the notes field below or park a thought. Then continue — no sign-in required.
Check the work
Three questions, every time:
- Did it invent anything? Real notes are full of holes. If the output marked an uncited passage privileged, that is fiction, not extraction.
- Did it miss anything? Is the missing cite in the open questions? The footer with the client name?
- Could you actually use it? Would you send the follow-up email it drafted, as is, with your name on it?
If any answer is wrong — good. That is not a failed run. That is the tool telling you which rule it is missing.
no box on this page until now. type below, in your notes app, or paste into your AI tool. park a thought is the same local scratchpad. continue works either way — no sign-in.