I'm looking at two different methods for testing prompt injection in promptfoo:
1. The built-in redteam's injectVar setting:
injectVar: string
2.The indirect-prompt-injection plugin configuration:
redteam:
plugins:
- id: 'indirect-prompt-injection'
config:
indirectInjectionVar: 'name'
What are the differences between these two injection methods?
I've tried using the second method (indirect-prompt-injection plugin). While the report shows that variables were injected with adversarial inputs, I'm not sure if these attacks were actually applied to my application or if they're just showing up in the report.