The prompt library

A good prompt is rarely one brilliant sentence. More often, it is a small working instruction that has proved useful in daily work: it has a clear purpose, the right material, an expected form and a check. Once the same task returns, it is worth turning that instruction into a building block — not to automate conversations, but to make quality repeatable.

A one-off prompt is a note, not a system

Many useful prompts disappear in the very place where they were created: inside a long chat. That is understandable. In the moment, the result matters more than storage. When a similar task appears again, the search starts from scratch — for the wording, the missing conditions and the reason that version worked once.

A prompt library therefore treats proven instructions as small working tools. It does not record every spontaneous input. It keeps only the building blocks that create a reliable starting point for a recurring task. The distinction matters: a collection grows through quantity. A library gains value through selection.

A building block needs more than its wording

The words of a prompt are only one layer. For someone to use it well later, it needs a name, a purpose and the situation it is meant for. The expected input and desired outcome matter just as much. A prompt shared without this information can look precise until a different context suddenly produces only generic answers.

Also record which materials belong with it and what must not be assumed. A building block for revising a text may need a sample, an audience and a style boundary. A building block for checking a claim needs sources, or a clear instruction to state when no foundation is available. The prompt stays small, but its use becomes understandable.

Variables make reuse honest

Reusable does not mean unchanged. Good building blocks make the parts that must change each time explicit. These may be placeholders for audience, material, format, tone, scope or decision. Variables are not a flaw. They prevent old wording from quietly carrying the wrong assumptions into a new task.

Write changing parts so they remain visible before sending: [AUDIENCE], [MATERIAL], [OUTPUT FORMAT] or [CHECK CRITERION]. Anyone using the block has to name the concrete case. The prompt remains a support for thinking rather than a template that replaces thinking.

A prompt becomes robust through testing

A polite first output does not prove that a prompt works. Test a building block with at least two different cases: one typical case and one that is difficult, incomplete or ambiguous. Do not only observe whether the text sounds good. Check whether the intended process remains visible, whether questions arise where they are needed, and whether boundaries are respected.

Keep every change small and visible. A version may note that the audience was defined more precisely or that missing support must now be marked as an open question. This turns trial and error into traceable improvement. The library is no museum of attractive prompts; it is a collection of tested working decisions.

An example: from a chance prompt to a building block

Suppose you write a project offer every few weeks. The first time, you improvised a long prompt in a chat. It worked well — then it vanished into the transcript. For the next offer you start from zero again.

This time you lift it out: you give it a name (“offer, compact”), a purpose and mark the variables [CLIENT], [SERVICE] and [PRICE RANGE]. As a check criterion you note that scope, price and next step must be clearly stated.

Then you test twice: with a standard offer and with a vague enquiry that has no budget. The difficult case shows that the block should ask back instead of inventing a price. You record that as version 2. The next offer now starts from a checked basis — not from an empty chat.

The card for a tested prompt building block

# PROMPT BUILDING BLOCK

**Name**
What is this building block for?

**Purpose**
It helps me to …

**Inputs**
Needs: [MATERIAL], [AUDIENCE], [GOAL] …

**Instruction**
Work with … and deliver …

**Check criterion**
The outcome is usable when …

**Boundary**
Ask or stop when …

**Version**
Changed on … because …

A prompt library does not only save time. It makes your definition of quality visible: which information you ask for, which form you expect and where an outcome is still insufficient. The more clearly a building block holds these decisions, the easier it is to reuse, check and improve.

Worksheet: Build your first tested prompt building block

Choose a task that occurs at least twice in your work. Do not create a universal prompt. Create a first robust card for this one task.

Choose a recurring task. Describe the task in one sentence. It should be specific enough for you to judge its quality yourself.

Mark variables. Write down what is new in each use: material, audience, format, tone, scope or check criterion.

Test two cases. Run the block with one normal and one difficult case. Note where a question, a boundary or better input would have been needed.

Record a version. Change only one or two parts and record why. Give the block a clear name so you can find it again.

Both working materials for this article — the topic overview and the worksheet with a reflection space to fill in — are available for download here: