Category
25 AI prompts for Coding
Review, refactoring, debugging and tests that actually catch things.
Written and tested for The AI University. Each carries its variables, a usage guide and the cases where it will let you down. Free with an account.
25 prompts
- Coding
Review for the Bug, Not the Style
A code review that hunts correctness failures and states the exact input that breaks each one, instead of listing naming preferences.
ClaudeChatGPTGemini000Open → - Coding
Failing Test First
Takes a bug report and writes the smallest test that fails for the right reason, before anything is fixed.
ClaudeChatGPTGemini000Open → - Coding
Refactor With a Safety Net
Plans a refactor as a sequence of individually safe, individually shippable steps, with the characterisation tests that make each one verifiable.
ClaudeChatGPT000Open → - Coding
Production Incident Post-Mortem
Turns an incident timeline into a blameless post-mortem that separates trigger from cause from contributing conditions, with actions that are actually assigned.
ClaudeChatGPT000Open → - Coding
Architecture Decision Record
Writes an ADR that captures why, including the options rejected and what would make you revisit it.
ClaudeChatGPT000Open → - Coding
API Design Review
Reviews an API for the mistakes that cannot be fixed later — naming, error semantics and anything that becomes a breaking change.
ClaudeChatGPTGemini000Open → - Coding
Database Migration Plan
Plans a schema change that ships without downtime, in steps that are each independently safe and reversible.
ClaudeChatGPT000Open → - Coding
Dependency Audit
Judges dependencies by maintenance risk and blast radius rather than by vulnerability count, which is what a scanner already tells you.
ClaudeChatGPT000Open → - Coding
Performance Profiling Plan
Sets a target and measures before optimising, so effort goes where the time actually is rather than where it feels slow.
ClaudeChatGPT000Open → - Coding
Error Handling Review
Finds every path that swallows an error, leaves state half-written, or reports success after failing.
ClaudeChatGPTGemini000Open → - Coding
Logging Strategy
Decides what to log from the questions you will need to answer at 3am, rather than logging everything and finding nothing.
ClaudeChatGPT000Open → - Coding
Caching Design
Decides invalidation before anything else, since a cache without an invalidation story is a bug with latency benefits.
ClaudeChatGPT000Open → - Coding
Feature Flag Plan
Plans a flag with a removal date, because flags are permanent by default and a codebase of them is unmaintainable.
ClaudeChatGPT000Open → - Coding
CI Pipeline Design
Optimises for the feedback loop, because a pipeline slow enough to work around is one people work around.
ClaudeChatGPT000Open → - Coding
Comment Pass
Removes comments restating the code and adds the ones explaining why, which is the only kind that does not go stale into a lie.
ClaudeChatGPTGemini000Open → - Coding
Type Modelling Review
Reshapes types so invalid states cannot be represented, rather than validating them at runtime and hoping.
ClaudeChatGPT000Open → - Coding
Concurrency Review
Names the shared mutable state first, because every concurrency bug in the code follows from it.
ClaudeChatGPT000Open → - Coding
Technical Debt Inventory
Ranks debt by the interest it actually charges per sprint, rather than by how much it offends whoever found it.
ClaudeChatGPT000Open → - Coding
Build vs Buy Analysis
Counts the maintenance decade rather than the build week, which is where build-versus-buy decisions actually go wrong.
ClaudeChatGPT000Open → - Coding
Observability Plan
Designs from the questions you will need to answer rather than from the metrics that happen to be available.
ClaudeChatGPT000Open → - Coding
Rate Limiting Design
Decides what happens to the rejected request, since a 429 with no guidance produces a client that retries into the wall.
ClaudeChatGPT000Open → - Coding
Backfill Script Review
Reviews a one-off data script for the three properties it needs — resumable, idempotent, rate-limited — before it is run at 2am.
ClaudeChatGPT000Open → - Coding
Module Boundary Review
Draws boundaries along what changes together, rather than along the organisation chart or the technical layer.
ClaudeChatGPT000Open → - Coding
Test Suite Triage
Finds the tests that cost more than they catch, and fixes the flakiness that teaches everyone to ignore a red build.
ClaudeChatGPT000Open →
24 of 25