What This Builds
A single Prompts table that serves your entire operating system. Every content type you generate, Substack Notes, blog posts, emails, social posts, selects its prompt from this one table, filtered by type. You will never build a second one. One brand or ten brands, one persona or twenty, the prompt library stays one table.
Why Build It
A prompt you type into a chat box works once and disappears. A prompt saved as a record works every time, for everyone on your team, and gets sharper with every batch you run.
The mistake operators make is scattering prompts everywhere: some in a doc, some in chat history, some rewritten from memory each time. Output quality swings wildly because the instructions swing wildly. A prompt library ends that. The same instructions run every run, improvements compound instead of evaporating, and anyone on your team generates on-brand content by selecting a prompt they did not have to write.
The reason it is one table and not one per brand or per content type: prompts are filtered at selection time by the Prompt Type field. Your Substack Note workflow only sees Substack prompts. Your future email workflow only sees email prompts. Splitting the table buys you nothing and costs you a maintenance headache for every workflow you wire later.
What You Need
Tools
Whatever database your system runs on. Airtable, Notion, a Google Sheet, or a table in your own app.
Setup Level
Ten fields, ten minutes. This is one of the smallest builds in the system and one of the highest leverage.
How It Works
Create one table named Prompts, then add these fields.
Name every prompt for the job it does, not the date you wrote it. "Contrarian Observation" tells you what you are selecting at generation time. "Note prompt v2" tells you nothing.
The category the prompt belongs to, and what filters the right prompts to the right workflow. Add options as you build: Substack, Blog Post, Social Content, Email Copy, Newsletter, Podcast, Video Script, General Content.
Use Active, Pending, and Archived. Pending is for prompts you are still testing. Archived keeps retired prompts for reference without cluttering working views. When you wire generation workflows, filter them to show Active prompts only.
The sub-format within the type. If Prompt Type is Social Content, the format might be LinkedIn or Instagram. If Prompt Type is Blog Post, the format might be listicle or how-to.
Cross-cutting labels for finding prompts later: hot take, extraction, launch, evergreen.
One or two sentences on when to reach for this prompt. Six months from now, you will not remember. Your team never knew.
The full prompt lives here, including the variable slot and the enforcement block.
When this prompt produces a piece you are genuinely proud of, paste it here. It becomes the quality bar and the fastest way to show a team member what this prompt is supposed to produce.
Link to your Brand table only when a prompt is built around one brand's specific positioning. Most prompts do not need it, and a blank Brand means the prompt works across the whole company. Your Brand record defines how you sound, the prompt defines how it is written.
Copy The Setup Prompt
Show the full setup prompt
I want to build a Prompt Library in [Notion / Airtable / Google Sheets / my own app — name your tool].
The goal is one Prompts table that serves my entire operating system. Every workflow should select prompts from this same table, filtered by Prompt Type and Status.
Help me set up a table named Prompts with these fields:
Name — single line text. Name each prompt for the job it does, not the date or version.
Prompt Type — single select. Start with: Substack, Blog Post, Social Content, Email Copy, Newsletter, Podcast, Video Script, General Content.
Status — single select. Options: Active, Pending, Archived. When I wire generation workflows later, I want them filtered to Active prompts only.
Content Format — single line text. The sub-format inside the prompt type, such as LinkedIn, Instagram, listicle, how-to, or extraction.
Tags — comma separated text or multi-select. Useful labels such as hot take, extraction, launch, evergreen.
Use Case Description — long text. One or two sentences explaining when to use this prompt.
Prompt Text — long text. The full prompt, including any variable slot and enforcement block.
Good Output Example — long text. A strong finished output from this prompt that can be used as a quality reference.
Brand — optional link to my Brand table. Only use this when a prompt is built around one brand's specific positioning. Most prompts should stay brand-blank so they can work across the company.
Then show me:
1. The table structure
2. The best views to create
3. How to add my first prompt record
4. How to refine a prompt when outputs miss
5. How to use Good Output Example as the quality bar over timeThe payoff: every content playbook you build from here forward starts with prompts already having a home. Build the table once, and the prompt section of any playbook is just adding records.