Create:Source Code AI prompts

Common prompts we can use with Duo Chat to be more efficient

From GitLab Discussion to Action

Use Case / Impact

Generating a detailed issue from a thread/discussion.

Impact: summarizing detailed discussions and capturing them in follow-up issues favour the iteration and time to merge of each individual step.

(See demo on levelup)

Prompt

Identify the discussion you want to use as source and copy the URL to the top comment of the thread (right-click on the timestamp and copy link).

Can you transform the following discussion TOP_COMMENT_LINK into an issue description? Keep it concise and useful for an engineer.

Security vulnerability detection in MRs

Use Case / Impact

When you’re working on an MR and want to check if you missed any security vulnerability, you can use this prompt to identify them.

Impact: This can lead to faster review times and decrease vulnerabilities shipped in our code.

(See demo on levelup)

Prompt

While on the Merge Request page, drop this prompt in the Agentic Duo Chat:

Can you detect any security concerns that are against OWASP's top 10 vulnerabilities in these MR changes? Provide examples.

Decoding Complex MR Feedback

Use Case / Impact

Sometimes you might get feedback that you don’t immediately and fully understand. This prompt can help decode and deconstruct that feedback to make it easier to understand and react to.

Impact: This can decrease review time by speeding up the discussion between author (you) and reviewer.

(See demo on levelup)

Prompt

  • Grab the link to the comment you want to check.
  • Write down your interpretation on what it means.

Use this information to compose a prompt such as:

From the reviewers comment LINK_TO_COMMENT, I understand the following YOUR_INTERPRETATION. Does this make sense? Am I missing something?

Debugging legacy code

Use Case / Impact

One of the most challenging things for our group (group::source code) is we work with legacy code that might be drifted quite apart from our current coding standards. More often than not, fixing something involves touching utils that are shared across the codebase. This broader reach means it takes a lot longer to verify a fix to avoid a regression.

Prompt

Can you give me an overview of the code logic for [COMPONENT_NAME]?
There is an potential xss issue, can you locate the problematic file/method?
Can you tell me where we use this method and which product feature it covers? 
Last modified September 19, 2025: Add Source Code AI Prompts page (db1924cb)