Unleash / unleash / 16726668527

86%

master: 91%

push

github

chore: AI flag cleanup should focus on relevant flag (#10462)

https://linear.app/unleash/issue/2-3749/prevent-ai-flag-cleanup-from-removing-the-wrong-flag

This slightly refines our prompt so we're strictly focusing on the
relevant flag, instead of mistakenly targeting other flags.

Also includes:
- Adding our prompt to our output so we can more easily debug it 
- Only grab the last cleanup summary in case there are multiple (more
than one step from the agent)
- Add a warning to the summary in case the agent couldn't find
conditional logic related to the flag, only definitions / configurations

Did a few manual tests and it seemed to work correctly. Example:
https://github.com/Unleash/unleash/pull/10461

7115 of 7167 branches covered (99.27%)

67239 of 73488 relevant lines covered (91.5%)

451.66 hits per line