Coding & building
Specify, test and debug small software changes with explicit checks.
You have code, an error, or an idea for a working tool.
Choose a guide by the result you want. Each one gives you a usable starting prompt, a concrete example and checks that expose weak output before you rely on it.
Can you reproduce the failure?
Share expected and observed behavior, the exact error and the smallest safe example. Ask for a minimal fix and one regression test.
Decisions that matter
- Remove credentials before sharing code.
- Test the original failure and an edge case.
- A plausible patch is not a passing test.
How to debug code with ChatGPT and check the fix
Show a small failing example, compare expected and actual behavior, and test a focused repair without hiding the error.
Field guideHow to review a codebase with ChatGPT without losing the real risk
Review a codebase with clear scope, trace real inputs, distinguish defects from guesses and verify a small repair with meaningful tests.
Field guideHow to protect secrets when asking ChatGPT about code
Build a useful debugging request with fake credentials, cleaned logs and enough structure to explain the failure.
Where the task can lead
How to write better ChatGPT prompts, with examples you can copy
Give ChatGPT a clear task, useful facts, and a format. Learn how to fix vague prompts and revise answers without adding a wall of instructions.
Connected guideWhat to do when ChatGPT does not follow your instructions
Find the exact mismatch, remove conflicting context, and repair the answer with a clear check instead of a longer prompt.
Bring the real material.
ChatGPT can work with a clear goal and relevant notes, files or constraints. Keep the source in view and inspect the final claim before using it.
See the first-session guide