ChatGPT Prompts for Software Engineers: 12 for Daily Work

ChatGPT prompts for software engineers save time on the writing and reasoning around code: design docs, reviews, tests, debugging notes, and status updates. Paste your own ticket, code, logs, or numbers, ask for a specific format, and check every suggestion against your codebase before you ship it. The 12 prompts below follow a feature from design to production.

Strip secrets and customer data before pasting anything: use placeholders for keys, tokens, and connection strings.

Design and plan

1. One-page design doc

From this ticket [paste] and these notes on our system [paste], write a one-page design doc: problem, goals and non-goals, options checked against [constraints], recommendation, rollout, and open questions.

2. Break a feature into tasks

Split this feature [describe] into tasks that each fit in [hours] and can be merged on their own. For each, give the files likely touched in [repo structure], the test to add, and what it depends on.

3. Trade-off check

Compare [option A] and [option B] for [problem] against these constraints [list]. Use only what I tell you about our stack [stack], and mark any claim about performance as needs measuring.

Write and review code

4. Self review before a PR

Review this diff [paste] as a strict reviewer on a [language] team. List bugs, missed edge cases, naming problems, and missing tests, ordered by severity, with the line each refers to.

5. Test cases from requirements

From these requirements [paste] and this function signature [paste], list unit test cases: normal, edge, and failure cases, with input and expected result for each. Then write them in [test framework].

6. Explain unfamiliar code

Explain what this code does [paste], step by step, for an engineer new to [codebase area]. Point out side effects, hidden assumptions, and anything that looks risky to change.

Debug and operate

7. Hypotheses from an error

Here is the error [paste], the stack trace [paste], and what changed recently [list]. Give me the three most likely causes, how to confirm or rule out each one, and the first command or log to check.

8. Incident timeline

From these alerts, chat messages, and deploy logs [paste], build an incident timeline with times in [time zone], the customer impact window, the trigger, and follow-up actions with owners. Do not guess times that are not in the logs.

9. Runbook entry

Write a runbook entry for [alert name]: what it means, how to check impact, the safe first steps, when to escalate to [team], and how to confirm it is resolved, based on these notes [paste].

Communicate clearly

10. PR description

Write a PR description for this diff [paste]: what changed, why, how to test it, risks, and screenshots or logs to attach. Keep it under [words] words.

11. Status update for non-engineers

Turn these engineering notes [paste] into a three-sentence update for [audience]: what is done, what is at risk, and what we need from them, with no jargon.

12. Code review reply

A reviewer left this comment [paste] on my code [paste]. Help me decide whether to accept it, push back, or ask a question, and draft a short, polite reply that explains my reasoning.

Write a design doc in one prompt

The flagship Software Engineer Prompt: Design Doc From Ticket and Notes takes your ticket, current system, constraints, options, traffic numbers, and team, and returns a one-page design doc with an options table, capacity math, a rollout plan, risks, and a test plan. Its example designs per API key rate limiting for 1,200 requests per second at peak.

Use AI on code safely

Follow your company's policy on which code and data can go into AI tools, and never paste secrets or customer data. Treat every answer as a draft from a fast reviewer: run the tests, measure performance yourself, and check suggested APIs against your library versions.

Related prompts

When something breaks, use the debugging prompt or browse ChatGPT prompts for debugging. For cleanup work, see the code refactoring prompt and ChatGPT prompts for code refactoring. For reviews, try the pull request code review assistant and the commit message and PR description prompt.

For agent setups, see the system prompt for a coding agent.

Find more in the Coding category, look through Browse, or post your own on Share.