Claude prompts are instructions tuned for Anthropic's Claude models, which excel at long-context reasoning, careful writing, and structured analysis. On Prompt Dock every Claude prompt is human-reviewed, and the in-app Playground runs against Claude directly — so you can reveal a prompt and immediately see how Claude responds. Whether you need a rigorous document summarizer, a coding assistant, or an agentic workflow, browse free and premium Claude prompts by category, each labelled with the model it was written for.
65 prompts
Give it a prompt you rely on and it builds the test cases that catch a regression when you edit it — normal inputs, edge cases, adversarial ones, and the pass criteria for each, so you can change a prompt without hoping.
Hand it an old function and your target language version, and get a modern rewrite that keeps the exact same behavior — with each change explained and every risky spot where old and new could diverge called out before you merge it.
Give it your topic, audience, and a few raw links or notes, and get a complete newsletter issue — subject line, opening hook, the main piece, short sections, and a closing CTA — written in one voice instead of assembled from fragments.
Paste docs nobody can follow and get them restructured around what the reader is trying to DO — with the prerequisites stated, the steps in order, and every unexplained assumption surfaced as a question you need to answer.
Paste a keyword list and get each one classified by intent, matched to the page type that can actually rank for it, and flagged where your existing page is the wrong format — the audit that explains why good content is not ranking.
Turn a raw incident timeline into a blameless postmortem with a real root cause, contributing factors, and action items that each have an owner and a trigger — the document a team can act on rather than file away.
Make the case for removing a feature — who still uses it, what it costs to keep, what breaks if it goes, and the migration path — so the decision gets made on evidence instead of deferred for another year.
Describe why customers churned and get a three-touch win-back campaign that addresses the actual reason they left — acknowledging it directly instead of pretending nothing happened, with a clear stop rule for who not to contact.
Name the assignment and what you are assessing and get a rubric with criteria described concretely enough that two graders agree — plus the feedback language that tells a student how to improve rather than just what they scored.
Paste how your team defines a metric and it finds the ambiguities that make two dashboards disagree — the unstated grain, timezone, filters, and edge cases — then rewrites it as a definition only one query can satisfy.
Describe an agent workflow and it maps where it will fail in production — loops, silent wrong answers, unsafe actions, and cost blowups — with the guardrail for each and the one that must exist before you ship.
Turn your quarter's raw numbers and events into a board update that leads with what changed, states the bad news plainly, and asks for the specific help you need — the update that earns trust instead of managing perception.
Paste a test that passes locally and fails in CI, plus the CI log, and get a ranked diagnosis of why it is flaky — timing, shared state, ordering, or environment — with the specific fix for each and a way to prove the flake is gone.
Paste a canned support reply and it rewrites it to sound like a person who read the ticket — keeping the accuracy, cutting the corporate padding, and marking exactly where the agent must personalize before sending.
Give it a product and get ten paid-ad headlines spread across five proven persuasion angles — outcome, pain, curiosity, social proof, and urgency. Enough range to find the one that clicks for your audience instead of betting on a single line.
Give a grade level and a topic and get an original passage of about two hundred words pitched precisely at that level, plus comprehension questions to match. Leveled reading material ready to hand out, without hunting for the right text.
Paste a wall of generic type errors and get the ONE real mismatch that caused the cascade, explained in plain terms, plus the minimal annotation that fixes it — without resorting to `any` or casting the problem away.
Paste a set of papers and get a structured synthesis — not a paper-by-paper summary, but four to six cross-cutting themes plus the open questions the field hasn't answered. The map that shows where the research agrees, conflicts, and still has room.
Describe a metric that suddenly moved and it walks the causes in the right order — instrumentation first, then mix shifts, then real behavior change — so you rule out the boring explanations before writing the story.
Paste this quarter's metrics and get a focused board-deck outline built to drive decisions — agenda, the story the numbers tell, and the specific asks. The structure that keeps a board meeting on the choices that matter instead of a data dump.
Paste interview transcripts and get the top five decision-ready insights — the things that recur or matter most — each backed by a real quote from the calls. Pulls the signal out of hours of conversation so the team acts on patterns, not the loudest anecdote.
List the emails you currently send and it audits the whole program — the gaps, the overlaps, the ones that fire at the wrong moment, and the ones to delete — so subscribers get fewer, better-timed messages.
Name a skill and get three fully stepped-out worked examples that build the procedure one move at a time, with the reasoning shown at each step. The scaffold that turns I-watched-it-done into I-can-do-it-myself.
Give a topic and a level and get a balanced ten-question quiz with a full answer key — a mix of multiple-choice and short-answer, calibrated to the learner. Ready-made assessment that checks real understanding instead of trivia recall.
Claude responds best to clearly structured instructions — explicit roles, XML-style sections, and long-context material. Claude prompts here are written and reviewed with those strengths in mind, and each names the Claude model it targets.
Yes — the in-app AI Playground executes prompts against Anthropic Claude models directly, so you see genuine Claude output before copying anything into your own tools. Every plan includes a daily token allowance.
Yes. Free Claude prompts can be revealed at no cost; premium ones unlock with an Explorer or Master subscription. All of them pass the same human review before publishing.