Organise your prompts with categories and tags
A simple categories-plus-tags system that scales from ten prompts to hundreds, mirrors the workspace sidebar, and keeps the right prompt one click away.
A prompt library starts tidy and ends as a junk drawer. The first ten prompts are easy to scan; by the fiftieth you’re scrolling, and by the hundredth you’re rewriting prompts you already own because you can’t find them. The fix isn’t more discipline - it’s a structure that does the remembering for you. This guide describes a two-layer system, categories plus tags, that scales without becoming a chore.
Categories are the shelves; tags are the labels
Think of a physical library. Categories are the shelves: broad, few in number, and every book lives on exactly one. Tags are the sticky labels you add: a book can have several, and they cut across shelves.
- A category answers “what kind of work is this for?” Each prompt belongs to one and only one. Keep the list short - six or so - so the choice is obvious and fast.
- A tag answers “what else is this about?” A prompt can carry many. Tags are how you capture the dimensions a single category can’t: the model it works best with, the project it belongs to, whether it’s a draft.
The discipline that makes this work is simple: one category per prompt, as many tags as are useful. If you ever feel a prompt needs two categories, that’s a sign your categories are too narrow - widen them and push the distinction down into tags.
A starter set of categories
The Promptsson workspace sidebar ships with six categories that cover most people’s work without overlap:
- Writing - drafting, editing, summarising, tone changes.
- Coding - code generation, review, debugging, explanations.
- SEO & Marketing - copy, headlines, keywords, campaign ideas.
- Research - questions, literature digests, comparisons, synthesis.
- Business - emails, plans, meeting notes, decisions.
- Learning - explanations, study aids, quizzes, practice.
These mirror the categories in the starter library, so a prompt you copy from the library lands somewhere sensible by default. Six is a deliberate number: broad enough that everything fits, narrow enough that you never hesitate over where something goes. Resist the urge to add a seventh until you’ve genuinely outgrown the six - most “new category” impulses are really tags in disguise.
Tags do the fine-grained work
Where categories are about the kind of work, tags describe everything else. Useful tag families include:
- Model:
gpt,claude,local- for prompts tuned to a particular model’s quirks. - Project or client:
acme,q3-launch,thesis- to gather everything for one effort. - Lifecycle:
draft,tested,favourite,archive- to separate the proven from the experimental. - Shape:
template,one-off,chain- how the prompt is meant to be used.
Keep tag names short, lowercase, and consistent. The single biggest tagging mistake is
near-duplicates - email, emails, and e-mail splitting one idea across three labels.
Pick a form and stick to it. If you only ever apply a tag to one prompt, it isn’t earning its
keep; a tag is useful precisely because it groups several things.
Make searching the default, browsing the fallback
A good structure means you rarely have to browse at all. Lead with search: type a few words
of what the prompt does and let the results narrow. Categories and tags then become filters
you stack on top - “Writing” plus favourite, or acme plus tested - to cut a long list
to a handful.
This is why naming matters more than filing. A prompt titled “Shorten a paragraph, keep the key warning” is findable by search no matter how it’s filed; one titled “v3 final” is lost the moment you forget the context. Title by the job the prompt does, file it under the obvious category, and add tags only where they’ll actually help you find it later. The advice in writing a reusable prompt about titling by job is the other half of this system.
Let the structure grow with you
Start minimal. Use the six categories and almost no tags. Add a tag the first time you think “I wish I could find all the prompts that are X” - that itch is the only good reason to create one. Periodically prune: delete tags you stopped using and merge near-duplicates. A library you trim once a month stays useful for years.
A few maintenance habits keep things healthy:
- Re-file on use. When you open a prompt and it feels mis-categorised, fix it then and there.
- Promote the proven. When a prompt graduates from
draftto reliable, retag it and, if it’s a template, make sure its variable names and field types are clear. - Prune deliberately. Retag prompts that still have a use, and delete stale ones after exporting a backup. A short rating note can record why a prompt fell out of favour.
Because everything lives on your own device - see local-first prompts - this structure is yours to shape with no limits and no one else’s conventions to fight.