Prompt library

Every prompt used in this book's playbooks, grouped by purpose. Swap the filenames, paths and fields for your own and they run.

The method behind writing them is in 08 Briefing it properly; this page is just the finished articles.

General clauses

These can be bolted onto any prompt.

Preview, then execute

Show me a preview first and execute only after I confirm.
Present the preview as a comparison table: current state / new state / basis.
Do not modify any file yet.

No invention

Where there is no clear basis, write "not specified" — do not infer.
Where the data does not support an inference, mark it "cannot be determined".

Specify the output

Output as a [Markdown / Word / Excel / PowerPoint / HTML] file saved to the
workspace, named [name].
No preamble and no closing summary boilerplate.

Have it ask you first

I want to do [thing]. What information do you need from me?
Do not start — list the inputs you need.

Self-check

Self-check before output: confirm the structure is complete, there is no
duplication, and the conclusions are concise.
List anything you are unsure about that I should verify.

Bound the change

Only change [specific part]. Leave everything else as it is.

File handling

Bulk rename

Read every file in the workspace and rename them in bulk following
"date + category + party + amount".

Rules:
- Date as YYYY-MM-DD, read from file content where possible, otherwise from
  the file creation time — and note which in the basis column
- Category from: travel / meals / office / other
- Party is the issuing organisation, abbreviated
- Amount as a whole number with the currency
- Keep the extension unchanged

Show me a preview first and execute only after I confirm.
Present it as a table: original name / new name / basis.
Do not modify any file yet.

File by content

Read every file in this folder, judge its type by content, and move it into
the matching subfolder: contracts / receipts / reports / images / other.

Create subfolders that do not exist.
List the moves for me to confirm, flagging the basis for each classification.
Do not move anything yet.

Generate an index

Create index.md in this folder with a table listing: filename / date /
category / party / key information. Sorted newest first.

Documents and decks

A formal document

Produce a [document type] covering:

1. [section one]
2. [section two]
3. [section three]

[Formal / concise / warm] register, for [audience].

Use [TBC] as a placeholder for anything undetermined — do not invent figures.
Output as a Word document saved to the workspace, named [name].docx.

From a template

[reference file] in the workspace is our standard format.

Follow its structure, heading hierarchy, register and typical length, using
the content in [new material] to write a new one.

Keep the same section order. Use the new data for the data sections and
rewrite the conclusions from it — do not carry over the original judgements.

Output as [filename].

A deck

Build a review deck from [source file] in the workspace.

Requirements:
- [N] slides maximum
- Clean and professional, for [setting]
- Foreground [point one] and [point two]
- Every slide title is a sentence stating a conclusion, not a noun phrase
- Data slides carry charts, not walls of text
- The final slide is an explicit ask

Output as pptx saved to the workspace.

Three versions of one document

From this content, produce three more versions:

1. A one-page summary (for decision makers, readable in 30 seconds)
2. A detailed version (for the delivery team, all details and assignments)
3. An external version (for partners, internal data and names removed)

Save as [name]-summary.md / [name]-detailed.docx / [name]-external.docx.

Outline first for long documents

Give me the outline first: section titles, one sentence per section on what it
resolves, and an estimated length. Do not write the body.

Meetings and minutes

Formal minutes

Read [log file] in the workspace and turn it into formal minutes:

1. Subject, time, attendees (as mentioned in the log)
2. Each agenda item: name → discussion points → conclusion
3. An action-item table: item / owner / due date / priority
4. Disputed or unresolved items in a separate section, noting the disagreement
5. Things needing follow-up that nobody claimed, listed separately

Rules:
- Where the log does not clearly state an owner or due date, write
  "not specified" — do not infer
- Conclusions must be ones actually reached; unresolved items go in section 4
- No preamble and no closing summary boilerplate

Output as Markdown saved to the workspace, named [name].md.

Transcription

Transcribe [audio file] in the workspace to [name].txt.

Mark speaker changes where you can (Speaker A, Speaker B is fine), and mark
anything unclear as [inaudible] rather than guessing.

From a group chat log

This chat log has a few decisions and to-dos scattered through it. Extract:

1. What was decided and who made the call
2. What to-dos exist, who claimed them, whether a time was mentioned
3. Questions raised that nobody responded to

In chronological order, noting the speaker of the original message for each.
Do not fill in anything that was not stated.

A pre-meeting brief

I have a "[subject]" meeting at [time].

Read [related files] in the workspace and produce a pre-meeting brief:

1. What was agreed last time and where each item stands now
2. The three changes in the data most worth discussing
3. A suggested agenda with a time estimate per item
4. Questions I should have answers ready for

One page maximum, saved as [name].md.

Data analysis

Round one: confirm understanding

Read [spreadsheet] in the workspace and answer only these four:

1. What fields does it have, and what do you understand each to mean?
2. How many rows? Multi-level headers or merged cells?
3. Which fields have missing values, and how many each?
4. Anything clearly anomalous — negative amounts, future dates, duplicate rows?

Answer only those four. Do not start analysing. Do not produce any files.

Round two: analyse and chart

Correct. Now:

Aggregate [metrics] by [dimension], producing:
- A bar chart: [description]
- A line chart: [description]

Then write a conclusion of under 250 words naming the two changes most worth
attention. For each: what happened, how large, what might explain it.

Where the data does not support an explanation, say "cannot be determined
from the data".

Output as an HTML page with interactive charts (values on hover), saved to
the workspace as [name].html.

Before merging tables

There are [N] monthly files in the workspace with supposedly identical field
structures.

Check first: are the field names identical? Does any have extra or missing
columns? Answer only that — do not merge yet.

Let it suggest the questions

Read this data and tell me: if you were our business lead, what three questions
would you most want answered, and why?

Do not analyse yet — questions and reasoning only.

Research

Time-bounded research

Search for significant developments in [field] over [time range] and summarise
the key points.

Requirements:
- Source link and publication date on every item
- Only items with concrete information: product releases, feature updates,
  funding, disclosed data, policy changes
- No opinion pieces, commentary or advertorial
- At most [N], ordered by importance
- One line on "who did what", then one on "why it matters"

Delete items where you cannot find a reliable source — do not keep them.

Competitor research

Research competitors [A, B, C] (site links below).

For each, establish:
1. Main features shipped in the last [N] months, in date order
2. Pricing (tiers, prices, core differences)
3. Publicly disclosed user scale or usage data
4. The differentiators they emphasise publicly

Data requirements:
- Source link and retrieval date on every item
- Distinguish "officially disclosed" from "third-party estimate / press report"
- Leave fields you cannot find blank and mark "no public information found"

Output as CSV first: vendor / category / content / source type / link / date.
I will decide about analysis after reviewing the data.

Cross-check a claim

For the claim "[claim]", find every source you can and compare what they say.

Where sources disagree, list them all with their sources.
Do not decide which is correct on my behalf.

Flag what needs verifying

Add a "needs human verification" section at the end listing every number, date
and attribution you are not fully confident about, and where you got each.

Content

Angles

I run a [positioning] account on [platform].
Tone: [three adjectives].
Readers: [age, occupation, context].

Based on [material], give me [N] content angles.

For each:
- Headline (under [N] words)
- One line on what the piece covers
- Why it appeals to my readers
- Likely format

Angles only for now — do not write the body.

The body

Take number [N], "[headline]".

Write it in full:
1. Headline (give me 3 options)
2. Body (under [N] words, in points, each with a concrete situation)
3. Hashtags (5–8, mixing broad and precise)
4. Cover copy (a main line and a sub-line)

[Register description].
Do not use [negative list].
Every point must have a real situation attached — do not just say it is useful.

A short-video script

Turn this into a [N]-second short-video script for [platform].

Structure:
- A 3-second hook (built on conflict or something counter-intuitive)
- Voiceover copy (timed by second, spoken register, short sentences)
- Shot suggestions (one line of visual per segment)
- A closing call to action

No "hi everyone, I'm X" opener.

Compliance check

Check this against Chinese advertising law and list every potentially
non-compliant phrase with an alternative.

Code and tools

A local tool

Build me a locally runnable [tool name].

[N] features only:
1. [feature one]
2. [feature two]

Requirements:
- Single-page app; double-clicking the HTML opens it, nothing to install
- Data stored in browser local storage, surviving a close and reopen
- Clean interface, [colour requirements]
- Works on a phone too

Build those and get it running. Do not add anything else.

Describing a bug

[Specific symptom, e.g. clicking save gives a white screen, console says TypeError].

To reproduce:
1. [step one]
2. [step two]

It does not fail when [the working case].

Have it explain a script

Explain what every line of this script does, marking specifically: which files
it reads, which it writes, whether it uses the network, whether it deletes anything.

Automation

Create a scheduled task

Create an automation that runs [frequency] at [time]:

[task description]

Save the result to [location] / push it to [channel].

A briefing

Generate a [content type] briefing and send it to [channel].

Sections:
1. [section one]
2. [section two]

Requirements:
- Every item must carry a source link; skip anything without a reliable source
- No preamble
- Subject line format: [Daily Briefing] YYYY-MM-DD
- Keep it to about three phone screens

Creating a Skill

Create a custom Skill called [name].

Trigger: when I [specific trigger].

It should:
1. [step one]
2. [step two]
3. Save to [path], appending to the end of the file, never overwriting
4. Reply with a confirmation naming [what to confirm]

Constraints:
- [constraint one]
- [constraint two]

Remote instructions

[Specific action], output to [specific path], and send me [key information]
when done.

If deletion or overwriting is involved, list the plan and wait for my confirmation.

---

Templates are a starting point

These were all extracted from runs that worked. Copy them the first time; after two runs you will certainly find a few places to adapt.

Save your adapted version, or better, turn it into a Skill.