Add AI skills to drive end-to-end tests#25354
Conversation
Generated by 🚫 Danger |
|
|
| App Name | Jetpack | |
| Configuration | Release-Alpha | |
| Build Number | 31375 | |
| Version | PR #25354 | |
| Bundle ID | com.jetpack.alpha | |
| Commit | 4ce4ab4 | |
| Installation URL | 6pjn4fq5kcv9g |
|
| App Name | WordPress | |
| Configuration | Release-Alpha | |
| Build Number | 31375 | |
| Version | PR #25354 | |
| Bundle ID | org.wordpress.alpha | |
| Commit | 4ce4ab4 | |
| Installation URL | 68ui02in8rn88 |
🤖 Build Failure AnalysisThis build has failures. Claude has analyzed them - check the build annotations for details. |
|
Hey Tony, is the plan to merge the PR or is it only for evaluation purposes? I've tried running it but quickly realized that it requires https://github.com/facebookarchive/WebDriverAgent, which I don't currently have installed. Would you mind sharing the setup steps required to get these skills to work please? What was your experience running |
|
@kean I think we can merge this PR. I'll ask the apps infra team to see if it's possible to run the tests on CI, which I'll leave them to explore now that a locally runnable pipeline is implemented in this PR. Have you tried the example prompts in the PR description? Do they work for you? Yes, WebDriveAgent is needed. But the skill should take care of that (cloning, running, etc), you should not need to do anything other than click "yes". |
I tried |
By "invoking WDA", do you mean making |
kean
left a comment
There was a problem hiding this comment.
The PR looks good. I replied to the RFC as well. Nice work 👏





Description
This PR adds two skills:
ios-sim-navigation. This skill is basically a reference of WDA, which Claude Code can use to interact with iOS simulators.ai-test-runner. This skill describes a workflow for running end-to-end tests on iOS simulators using theios-sim-navigationskill.You can use
ios-sim-navigationto verify your code changes. For example, you can add this to your prompt:You can try the
ai-test-runnerlocally using a prompt like this:I have translated some UI tests into test cases in
Tests/AgentTests/ui-tests. If you want, you can ask Claude Code to run the full test suite, which takes 40+ minutes. You can also ask Claude Code to run on two simulators simultaneously, which halves the runtime.