10 AI Code Review Prompts I Actually Use

I review 15-20 PRs per week. These are the prompts that survived — the ones I keep coming back to because they consistently catch real issues.

Copy-paste them into Claude Code, Copilot Chat, Cursor, or any LLM CLI.


1. Bug Scan

S-----FcoaNRRSErnueaQrmlscLratleothuir:iurcnsncojhfdeneaidedcnliflitdefietil,fnaioiekonnlfdsngiosonrr(rge:iuoa,snrXpkcSssslcSeoovsnveecurdulirntfreyierlna(ecbHsyiI,lGiiHcst/osiMnueEnesDes/cLtOiWo)n,s,exsptlraenaamtsi)on.

2. Security Review

R-----IegvIASIIninuennoeptnjsrwuhseeeteicctnttushvtiirtiaivoeyslcenliadecdtdveoaiaufidtotlaseinanuso/elucnaertehuxassagtpb.nahoigposlOesruinirtlfzeiyoaert(sfillsoo(aengSgcsQub,Ltry,hipeitarcnysrogsomsiermssatsmnhuedaes,tssaLcogDoneAulsPly,)d:rbeespeoxnpsleosi)ted.

3. Style & Pattern Check

R-----DeovFIMCAniunion'encsdttwcoseitni-ftisntplhoighaainsatgsstotteremcndriovtmunndeipsoernslralisf5mecpso0iaaetrndtcy:lgieilinsfencgieoecpsndxrvoitetecsofhnsteatti[rtirnPeoigynsnntchsguheostosuin.ll/diJOtanbivleeays/sTfpyllpaiegtSctrhiipntg]sthataffectreadabilityormaintainability.

4. Test Coverage Gaps

GF----FioovcEECSreudrotnsgrnaeeoctatoruechncrhi:aprtssaermetnaiishcnsmsyssp(iile(stnemncigmpeeoettnntnywasetorsairittnko,ipsoufentax,ipcllhnuaaurinlenglse,,w,hbtawoithumanectdooauutrtleysds,tvgaioclnauvwseareslos)indgardieantmapi)rsosdiuncgt?ionwithoutit.

5. Performance Review

R-----EesvNMUBMti+inliie1snosmwsecsaqickittuneinehegsngirsgisyiapmnrcapcpdyagaoaelicdtxmlnteteesa:esmtfroiiwon(rnoirsbynlaalpsasoeelyrtrdlnhfocuioocnsrnacbmtoomaqinuanuotntcenedtersxeeytdrisspqasaututeetr1sei0:re0nssr)eq/s?10,000req/s?

6. API Contract Check

R-----AesvBMIMUsirininueescsdmwasosoekinictinsnuehngigmxigseisitonscnerttAhpneiPautidnIntnggecscevroichsarrdlaloreintireegodceneartftetefsfxisHeoiopTcdrsnoTte:tnPspisenesngtdfacotolrunimseatntchtsoesdecsurrentcontract.

7. Database Migration Review

R-----evWIDACiisoroeleeuwlislttdtiihhtrteiiertslvheoeafdcrnmaaksdiitilslatbesbalimabeenisl?xgdeei-sIsiwmstnaididygutnerrhgxaaieentndsdigdaootflnewaoexnrafegcvocroneruare:tdrwtieehoccencto?ollsryucFr?mhoenercsmtah?uosiwendlaoinnbgr?WoHkEeRnEsctlaatues?es?

8. Error Handling Audit

T-----raWADAIchrorseaeeteteerthvhrrheearoerrporreyprseesrenpcerslratrootrowgpcyrhgaheeg-lpndaaoatlgtewelihaicctchihhoawnrnheerdetxnelrhtoceeieutrsrglstnhyhcateolcthrdooaeecnt:attslehshlxewotaufcllataldiolollwbsdee?eri?b?mupIgos?rtiatntboeurnrdoerds??

9. “What Would Break” Analysis

AF----sosrWWHHuhhoomeaawweattcyythttooh:rhuuiie''sgddgucesdforeeidsrtxe/eisciityttsstidetemp(limomoypneaidcttotrowionpugrl/odadlubecerttiso)n.Listthetop5thingsmostlikelytogowrong,rankedbyprobability.

10. Onboarding Readability Check

R----BeeaWWWAdhhhrhaeaeottrtnh'eveiscasswortcomicoumaonlebdfdnlieutefsys/aiofssnuouogrnmincefoedttreoihydcoiouunnuntmg'coenrlnareeatmeasaeqarktsu?iinsosreoneewmlsewfeoo-tnnuergelxiidpbnflaeohalerenrlakcptnjo?ooonrwitylne?eixdntgg?etthoeutnedaemr.stand,flagit.

How I Use These

I don’t run all 10 on every PR. My typical flow:

  • Every PR: #1 (Bug Scan) + #3 (Style)
  • Security-sensitive PRs: Add #2
  • Database changes: Add #7
  • New features: Add #4 (Test Gaps) + #9 (What Would Break)
  • Performance-critical: Add #5

Want to see how I use these in a full workflow? Read my AI code review guide.


Want more templates like this? Subscribe to my newsletter → — I send one practical AI workflow per week, including prompt templates I don’t publish on the blog.