Activation criteria extend to general research questions and incidental fact-checking without clear invocation boundaries.
- Category
- Trigger Abuse
- Confidence
- 88% confidence
- Finding
Whether a multi-model conclusion 'would help' is subjective, and needing to check a factual claim can occur during almost any task. These conditions do not distinguish a user-requested external research workflow from ordinary answering or internal verification, potentially causing unintended KARP invocations.
- Content
