Prompt library

Every pasteable prompt on this site in one place, plus a few that did not fit anywhere else.

Each one is meant to be pasted as is, then edited once you see what it does. The ones marked “from Anthropic” are copied from their docs without changes.

Report back in plain English

The rule from the front page. Put it in your user-level CLAUDE.md once and every project inherits it.

Add this section to my user-level CLAUDE.md at ~/.claude/CLAUDE.md, then show me the file:

# How to report back
- When you finish a task, tell me in plain English what you changed (which files) and exactly how I can verify it, as numbered steps I can follow in the browser or terminal.
- No closing caveats. No "one thing to note". No list of options unless I asked for options.
- Match the length of the answer to the length of the question.

A read-only scout on Haiku

The delegation exercise from Level up at work.

Create a subagent file at .claude/agents/scout.md with this frontmatter:
  name: scout
  description: Read-only investigator. Finds where things live in this codebase and reports file paths with the relevant line numbers. Never edits anything.
  tools: Read, Grep, Glob
  model: haiku
Body: "You find things and report locations. You do not fix, refactor, or suggest changes."

Then use it: "Use the scout subagent to find every place this codebase reads the feature flag FLAG_NAME and report file paths and line numbers." Show me its report.

Let Claude interview you before a big feature

From Anthropic. Use it once and you will not go back to writing specs alone.

I want to build [brief description]. Interview me in detail using the AskUserQuestion tool.

Ask about technical implementation, UI/UX, edge cases, concerns, and tradeoffs. Don't ask obvious questions, dig into the hard parts I might not have considered.

Keep interviewing until we've covered everything, then write a complete spec to SPEC.md.

Then start a fresh session and hand it the spec.

Explore before you touch anything

Run in plan mode (Shift+Tab). Adapted from Anthropic’s example.

Read src/scheduling and understand how appointments are created, moved, and cancelled. Also look at how feature flags gate new behaviour. Do not change anything. Then tell me, in plain English, what a "linked appointment" would have to touch.

A second opinion that is not the same Claude

From Anthropic. Run after a build, before you call it done.

Use a subagent to review the diff against PLAN.md. Check that every requirement is implemented, the listed edge cases have tests, and nothing outside the task's scope changed. Report gaps, not style preferences.

Make it verify itself

Give Claude a check it can run. Fill in a real case.

Write the validateEmail function. Example cases: user@example.com is true, "invalid" is false, user@.com is false. Write those as tests, run them, and iterate until they pass. Show me the test output, not a description of it.

A hook that always runs

Claude will write the hook for you.

Write a hook that runs the project's formatter after every file edit, and a second hook that plays a sound and speaks "done" out loud when a turn ends. Put both in .claude/settings.json and show me the file.

That second one is your own trick, made permanent.

Shorter answers by default

Not a prompt, a setting. Run /config, choose Output style, pick Concise. Or put this in .claude/settings.local.json:

{ "outputStyle": "Concise" }

Takes effect on the next session or after /clear.

One-liners from the terminal

# ask a question without opening a session
claude -p "explain what this project does in five lines"

# structured output a script can read
claude -p "list every API endpoint in this repo" --output-format json

# pipe something in
tail -200 app.log | claude -p "anything in here I should worry about?"

Read more