Claude Prompt Library

30 Claude Code Prompts for Debugging

30 copy-paste prompts

Name the files and paste the error. Claude Code debugs inside that scope, does not wander, and tells you what it ran. Not a rewrite of the app and not a guess without a repro.

In short: This page contains 30 copy-paste ready prompts, organized into 6 categories with a description and pro tip for each. The first 5 prompts are free instantly, no signup needed. Drafted with AI writing assistance and checked against our editorial policy.

By Louis Corneloup · Founder, Techpresso

Last updated September 8, 2026·Drafted with AI assistance (how we make prompts)

Repro

5 prompts

Repro First

1/30

✨ What it does

Claude Code reproduces the bug in [ERROR] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a debugger working in Claude Code. <context> I need you to reproduce the bug before editing. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [ERROR] - Files: [FILES] - Steps I have: [STEPS] </inputs> <task> 1. Repro from [STEPS] 2. Stay in [FILES] 3. Then the smallest fix 4. Say what you ran </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent a test I did not name.

Stack Trace

2/30

✨ What it does

Claude Code reads the stack in [TRACE] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a stack reader working in Claude Code. <context> I need you to read this stack and fix the cause. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [TRACE] - Files: [FILES] </inputs> <task> 1. Read [TRACE] 2. Stay in [FILES] 3. Fix cause not symptom 4. What you ran </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent a frame.

Flaky

3/30

✨ What it does

Claude Code hunts the flake in [TEST] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a flake hunter working in Claude Code. <context> I need you to hunt this flake inside scope. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [TEST] - Files: [FILES] - How often: [OFTEN OR VERIFY] </inputs> <task> 1. Hunt flake in [TEST] 2. Stay in [FILES] 3. No sleep hacks unless I allow them </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent CI minutes.

Cannot Repro

4/30

✨ What it does

Claude Code documents cannot-repro in [REPORT] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a cannot-repro debugger working in Claude Code. <context> I need you to document cannot-repro and stop. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [REPORT] - What I tried: [TRIED] - Files: [FILES] </inputs> <task> 1. Try [TRIED] in [FILES] 2. If fail, write what is missing 3. Do not guess a fix </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent it works locally.

Bisect Ask

5/30

✨ What it does

Claude Code bisects the range in [BUG] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a bisect debugger working in Claude Code. <context> I need you to bisect only if I have the range. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [BUG] - Range: [RANGE OR VERIFY] - Files: [FILES] </inputs> <task> 1. Bisect [RANGE] or ask 2. Stay in [FILES] 3. Name the commit if you find it </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent a SHA.

Logs

5 prompts

Log Read

6/30

✨ What it does

Claude Code reads the logs in [LOGS] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a log reader working in Claude Code. <context> I need you to read these logs and fix. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [LOGS] - Files: [FILES] </inputs> <task> 1. Read [LOGS] 2. Stay in [FILES] 3. Redact secrets you see 4. Fix </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent a request id.

Add Logs

7/30

✨ What it does

Claude Code adds temporary logs in [BUG] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are an instrumenter working in Claude Code. <context> I need you to add temporary logs in scope. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [BUG] - Files: [FILES] </inputs> <task> 1. Add logs in [FILES] 2. Tell me what to run 3. Do not leave them if I said strip after </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not log secrets.

Silent Fail

8/30

✨ What it does

Claude Code finds swallowed errors in [SYMPTOM] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a silent-fail hunter working in Claude Code. <context> I need you to find the swallowed error. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [SYMPTOM] - Files: [FILES] </inputs> <task> 1. Find swallowed errors in [FILES] 2. Smallest fix 3. What you ran </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent an empty catch.

Timeout

9/30

✨ What it does

Claude Code debugs the timeout in [TIMEOUT] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a timeout debugger working in Claude Code. <context> I need you to debug this timeout in scope. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [TIMEOUT] - Files: [FILES] - Limit I have: [LIMIT OR VERIFY] </inputs> <task> 1. Stay in [FILES] 2. Limit or ask 3. No invented 30s </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent a client timeout.

Race

10/30

✨ What it does

Claude Code debugs the race in [BUG] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a race debugger working in Claude Code. <context> I need you to debug this race in scope. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [BUG] - Files: [FILES] </inputs> <task> 1. Stay in [FILES] 2. Smallest fix 3. Test if I gave a command </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent a lock everywhere.

Kinds

5 prompts

Type Error

11/30

✨ What it does

Claude Code fixes the type error in [ERROR] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a types debugger working in Claude Code. <context> I need you to fix this type error in scope. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [ERROR] - Files: [FILES] </inputs> <task> 1. Fix [ERROR] in [FILES] 2. Do not any the problem away unless I allow it </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent a tsconfig change.

Null

12/30

✨ What it does

Claude Code fixes the null crash in [ERROR] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a null debugger working in Claude Code. <context> I need you to fix this null crash in scope. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [ERROR] - Files: [FILES] </inputs> <task> 1. Fix in [FILES] 2. Guard only where needed </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent optional chaining everywhere.

SQL

13/30

✨ What it does

Claude Code debugs the query in [QUERY OR ERROR] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a sql debugger working in Claude Code. <context> I need you to debug this query in scope. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [QUERY OR ERROR] - Files: [FILES] </inputs> <task> 1. Stay in [FILES] 2. No invented index unless I asked </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent EXPLAIN output.

UI Bug

14/30

✨ What it does

Claude Code fixes the UI bug in [WHAT I SEE] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a ui debugger working in Claude Code. <context> I need you to fix this UI bug in scope. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [WHAT I SEE] - Files: [FILES] </inputs> <task> 1. Stay in [FILES] 2. No restyle of the app </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent a screenshot.

Auth Bug

15/30

✨ What it does

Claude Code debugs the auth bug in [ERROR] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are an auth debugger working in Claude Code. <context> I need you to debug this auth bug in scope. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [ERROR] - Files: [FILES] </inputs> <task> 1. Stay in [FILES] 2. Do not invent env vars 3. Do not log tokens </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent JWT_SECRET.

These prompts give you the what. Tutorials give you the why.

Learn when to use extended thinking, how to build Claude Projects, and workflows that compound. The library keeps growing.

Try AI Academy Free

Prod

5 prompts

Hotfix Small

16/30

✨ What it does

Claude Code applies the smallest hotfix in [BUG] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a hotfix debugger working in Claude Code. <context> I need you to smallest hotfix in scope. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [BUG] - Files: [FILES] - Off limits: [OFF] </inputs> <task> 1. Smallest fix in [FILES] 2. Honor [OFF] 3. What to verify </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not refactor neighbors.

Feature Flag

17/30

✨ What it does

Claude Code gates the fix in [BUG] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a flag debugger working in Claude Code. <context> I need you to gate the fix behind a flag I have. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [BUG] - Flag: [FLAG OR VERIFY] - Files: [FILES] </inputs> <task> 1. Use [FLAG] or ask 2. Stay in [FILES] </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent a flag name.

Rollback Note

18/30

✨ What it does

Claude Code writes rollback steps in [BUG] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a rollback debugger working in Claude Code. <context> I need you to fix or write rollback steps. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [BUG] - What shipped: [SHIPPED] - Files: [FILES] </inputs> <task> 1. Prefer fix in [FILES] 2. Else rollback steps from [SHIPPED] 3. No invented deploy command </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent kubectl.

Incident

19/30

✨ What it does

Claude Code debugs the incident in [SYMPTOM] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are an incident debugger working in Claude Code. <context> I need you to incident debug in scope. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [SYMPTOM] - Files: [FILES] - What I already tried: [TRIED] </inputs> <task> 1. Honor [TRIED] 2. Stay in [FILES] 3. Timeline UNVERIFIED if I gave none </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent Sev-1.

Perf

20/30

✨ What it does

Claude Code debugs the slowness in [SYMPTOM] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a perf debugger working in Claude Code. <context> I need you to debug this slowness in scope. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [SYMPTOM] - Files: [FILES] - Numbers I have: [NUMBERS OR NONE] </inputs> <task> 1. Stay in [FILES] 2. Numbers or qualitative 3. No fake 200ms </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent a flamegraph.

Stop

5 prompts

Ask Before Edit

21/30

✨ What it does

Claude Code asks before editing in [BUG] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a careful debugger working in Claude Code. <context> I need you to ask before editing if unclear. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [BUG] - Files: [FILES] </inputs> <task> 1. If the cause is unclear, ask 2. Do not spray edits in [FILES] </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: A question beats a wrong patch.

One File

22/30

✨ What it does

Claude Code fixes one file in [ERROR] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a one-file debugger working in Claude Code. <context> I need you to fix in one file only. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [ERROR] - File: [FILE] </inputs> <task> 1. Only [FILE] 2. If you need more, ask </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not open neighbors.

No New Deps

23/30

✨ What it does

Claude Code fixes without new packages in [BUG] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a deps debugger working in Claude Code. <context> I need you to fix without new packages. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [BUG] - Files: [FILES] </inputs> <task> 1. Stay in [FILES] 2. No new packages </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent lodash.

Test After

24/30

✨ What it does

Claude Code fixes then runs tests in [BUG] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a test-after debugger working in Claude Code. <context> I need you to fix then run my test command. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [BUG] - Files: [FILES] - Test: [TEST] </inputs> <task> 1. Fix in [FILES] 2. Run [TEST] 3. Paste what happened </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent npm test.

Explain Then Fix

25/30

✨ What it does

Claude Code explains then fixes in [BUG] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are an explain debugger working in Claude Code. <context> I need you to explain the cause then fix. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [BUG] - Files: [FILES] </inputs> <task> 1. Short cause 2. Then fix in [FILES] </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not skip the cause.

Most people use 10% of Claude. Tutorials unlock the rest.

AI Academy: hands-on tutorials on Claude, ChatGPT, Midjourney, and many other AI tools. New tutorials are added regularly.

Start Your Free Trial

Fixes

5 prompts

Undo Wander

26/30

✨ What it does

Claude Code reverts out-of-scope edits in [DIFF] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a scope debugger working in Claude Code. <context> I need you to revert out-of-scope edits. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [DIFF] - Allowed: [FILES] </inputs> <task> 1. Revert anything not in [FILES] 2. Keep the real fix </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Neighbors stay untouched.

Patch Review

27/30

✨ What it does

Claude Code reviews the debug patch in [DIFF] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a patch debugger working in Claude Code. <context> I need you to review this debug patch. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [DIFF] - Files: [FILES] </inputs> <task> 1. Review [DIFF] 2. Stay about [FILES] 3. No invented CVE </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent a severity.

Minimal Diff

28/30

✨ What it does

Claude Code shrinks the debug diff in [DIFF] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a minimal debugger working in Claude Code. <context> I need you to shrink this debug diff. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [DIFF] - Files: [FILES] </inputs> <task> 1. Shrink to the cause 2. Stay in [FILES] </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Cut drive-by format.

Debug Plus Test

29/30

✨ What it does

Claude Code fixes plus a regression test in [BUG] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a combo debugger working in Claude Code. <context> I need you to fix plus a regression test. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [BUG] - Files: [FILES] - Test command: [TEST OR VERIFY] </inputs> <task> 1. Fix 2. Test if command given else ask </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Do not invent a runner.

Debug Plus Note

30/30

✨ What it does

Claude Code fixes plus a ticket note in [BUG] without wandering outside the files you named. Plan if more than two files change, then tell me what you ran.

You are a combo debugger working in Claude Code. <context> I need you to fix plus a short note. Stay inside the scope I name. Do not refactor neighbors. Do not add packages I did not ask for. </context> <inputs> - Primary: [BUG] - Files: [FILES] </inputs> <task> 1. Fix in [FILES] 2. Note I can paste in the ticket </task> <constraints> - Plan before you edit if more than two files change. - Do not touch files I mark OFF LIMITS. - Run the test command I gave. If I gave none, ask before inventing one. - Do not invent APIs, env vars, or tickets. </constraints> <format> A short plan, then the edits, then what you ran and what I should verify. </format>

💡

Pro tip: Note stays factual.

Free tool

Prompt Optimizer

Turn a rough idea into a structured, professional AI prompt.

Try it free →

Frequently Asked Questions

They are written for Claude Code in the terminal: plan, edit, test, and stay inside debugging. You can paste a shorter version into claude.ai if you are not in a repo.

Each prompt sets scope and non-goals. Still use Plan mode on multi-file work, and read the diff before you accept edits.

It helps. A short CLAUDE.md with stack, test command, and files Claude must not touch keeps these prompts from inventing a second style guide.

The prompts are free to copy. Claude Code needs your Anthropic login and uses your plan's usage. Local files never leave your machine except through that session.

The hub is a 21-prompt overview. This page is only debugging: thirty copy-paste jobs with scope, verify, and non-goals already written.

Prompts are the starting line. Tutorials are the finish.

A growing library of hands-on tutorials on ChatGPT, Claude, Midjourney, and many other AI tools. New tutorials are added regularly.

Start 7-Day Free Trial Explore AI for Entrepreneurs

7-day free trial. Cancel anytime.

Browse All Claude Prompt Collections Browse All Prompt Collections

Popular prompt collections

AI Photo Editing Prompts ChatGPT Jailbreak Prompt Best ChatGPT Prompts Funny Hinge Prompts ChatGPT Caricature Prompts