Guides

From a shared prompt doc to a team skill library

A shared prompt document (in Notion, Google Docs, or a wiki) is the natural first version of a team prompt library, and it fails in a predictable way: every use requires someone to find the prompt, copy it, and paste it, so the document and reality drift apart. The replacement is a skill library, where the same instructions load into your team's AI tools automatically and nobody pastes anything.

What the prompt doc gets right

It centralizes knowledge, it needs no setup, and anyone can edit it. Those are real virtues, and for a handful of occasionally-used prompts a doc is genuinely enough.

Where it breaks

The paste tax. Every use is a retrieval chore, so people stop retrieving and work from memory. No delivery. When the doc changes, nothing already pasted anywhere updates; the doc is upstream of nothing. No relevance matching. A doc cannot tell an AI tool "use this one when someone asks about refunds"; a skill's description does exactly that, which is why skills fire without being asked for. No history that means anything. Doc version history shows edits, but there is no concept of "the version everyone's AI is currently using."

What a skill library changes

Skills are the same content the doc held, in the standard SKILL.md format (what a skill is), with a name and a when-to-use-this description. They load into Claude, ChatGPT, Cursor, and other tools automatically (how cross-tool delivery works), so "using the prompt library" stops being an action anyone takes. Editing stays as easy as the doc: say what is wrong in chat. What you gain over the doc is delivery, relevance matching, attributed version history, and one-sentence revert.

Migrating

Migration is conversational: paste a prompt from the doc into a chat and say "save this as a team skill for X." Ten prompts take an afternoon. Keep the doc during the transition; retire it when people notice they have stopped opening it.

SkillRelay is a hosted skill library that syncs across the AI tools a team uses. Overview · Product reference · How it works