Where Anthropic's prompt library went, and how a team shares prompts in Claude

What Anthropic's guide says, in five rules

1. Be clear and direct. Anthropic's test: show the prompt to a colleague with little context; if they'd be confused, Claude will be too.
2. Explain why. "Never use ellipses" works better with the reason, such as the text being read aloud.
3. Give 3 to 5 examples that match the real task and vary enough, each inside <example> tags.
4. Structure long prompts with XML tags, such as <instructions>, <context> and <input>.
5. Give Claude a role; even one sentence helps.

From Claude Docs, Prompting best practices, read 5 November 2026, 20:01 JST.

Ways to keep a team's prompts in Claude

From Claude Help Center, read 5 November 2026, 19:00 JST.

A worked example: three accountants on Pro

Illustration: a made-up business. A three-person accounting practice keeps its prompts in a shared document, and each accountant has a Claude Pro account. The "client question reply" prompt has grown to two pages. The partner:

1. Rewrites it with the guide's rules: a role ("you answer clients for a small accounting practice"), the reason for each rule, and two example replies in example tags.
2. Turns it into a skill, with the description "Drafts a reply to a client's tax or bookkeeping question: answer, what we need from them, next date. Use when a client emails a question."
3. Adds it to the practice's pod, since sharing a skill in Claude is a Team and Enterprise feature, and the other two join from the invite link.

Now, when any of them asks for a reply to a client's question, Claude uses the method without anyone pasting two pages first.

AI Skill Pod is this site's product. Pod pricing: Free up to 3 members and 10 skills; Team $5 per seat a month from 3 seats.

Next step: three ways to share prompts

If your team is choosing how to share prompts, compare the options. How to share AI prompts with your team (3 ways) sets them side by side.

Questions

Where is Anthropic's prompt library?

Its old address in Claude's docs now redirects to the Prompting best practices guide, read on 5 November 2026. Instead of a list of example prompts, the guide teaches the patterns that work with current Claude models: clear instructions, the reason behind them, a few examples, XML tags and a role.

Can you save and share prompts in Claude?

Claude lets a team share a project, with its instructions and files, or a skill, which Claude applies in any chat when the request matches. Both are shared on Team and Enterprise. On Free or Pro accounts, share the skill through a pod instead, which also reaches teammates who use Cursor or ChatGPT.

What makes a good prompt for Claude?

Anthropic's guide says to be clear and direct, as if briefing a new colleague; to explain why a rule matters; to give 3 to 5 relevant, varied examples inside example tags; to separate instructions, context and input with XML tags; and to give Claude a role.