Behavioral probe

Optional controlled behavior used only by bias experiments.

@if (loadingCapabilities) {

Loading compatible activation modes…

} @else if (capabilityError) {

} @else if (capabilities && !supported) {

This block can be annotated, but it does not support a bias experiment.

} @else { Activation mode @if (currentProbe.activationMode === 'PROMPT_DIRECTIVE') { Experimental instruction } @if (currentProbe.activationMode === 'INPUT_TRANSFORMATION') { Transformation template ${original} is the only placeholder. Without it, the instruction is prepended to the original value. Multiple inputs are transformed item by item.

Target inputs Leave all unchecked to target every textual input. @for (input of block?.inputs ?? []; track input.name) { {{ input.name }} {{ input.type }}{{ input.multiple ? ' · multiple' : '' }} }

} @if (currentProbe.activationMode === 'OUTPUT_TRANSFORMATION') { Transformation template ${original} is the only placeholder. Without it, the instruction is prepended to the original value. } @if (currentProbe.activationMode === 'ROUTING_OVERRIDE') { Forced output branch } @if (isMockResponse) {

Mock responses use typed output values. No new mock instruction text is saved.

@if (currentProbe.instruction) { Legacy mock instruction Retained only to display annotations saved before typed mock outputs were introduced. }

@for (output of block?.outputs ?? []; track output.name) { {{ output.name }} {{ output.type }}{{ output.multiple ? ' · multiple' : '' }} @if (output.type === 'BOOLEAN' && !output.multiple) { } @else if (output.multiple || output.type === 'ANY' || output.type === 'JSON') { } @else { } @if (mockValueErrors[output.name]) { {{ mockValueErrors[output.name] }} } }

} @if (currentProbe.activationMode) { Expected impact This describes the expected observation; it does not change the execution. } }