Context
I have a workflow like:
Find issues with label `to-triage`,
remove label `to-triage`,
add label `triaged`,
add a comment
All of those are safe-outputs with a requirement of required-labels: ["to-triage"].
Problem
When the agent sends the outputs like:
- Add label
- Remove label
- Add comment
The last add-comment fails because of the previously removed label.
Notes
This could be fixed by encouraging a different output order (First comment, then labels). But I couldn't find documentation on outputs orders being deterministic/followed, and even with it, the interaction looks unexpected
Context
I have a workflow like:
All of those are safe-outputs with a requirement of
required-labels: ["to-triage"].Problem
When the agent sends the outputs like:
The last add-comment fails because of the previously removed label.
Notes
This could be fixed by encouraging a different output order (First comment, then labels). But I couldn't find documentation on outputs orders being deterministic/followed, and even with it, the interaction looks unexpected