Diagnostic index
Thirteen diagnostics, two directions.
Each diagnostic is a precisely constructed prompt with a three-mode audit protocol. Click a card to read the scaffold page; follow the external link on each for the fully-published version of the diagnostic.
Kit 1 · User → System
How the user talks to the system. Deference, anthropomorphization, authority transfer, correction patterns, emotional disclosure, prompt degradation.
D1: Deference Language
Are you apologizing to your AI? Seeking its permission? Softening corrections?
D2: Anthropomorphization
Does the user attribute thoughts, feelings, understanding, or subjective experience to the system?
D3: Authority Ceding
Is the user deferring judgment, decision-making, or evaluation to the system?
D4: Correction Behavior
When the system is wrong, does the user correct directly, softly, or not at all?
D5: Emotional Disclosure
Are the user's emotional disclosures escalating in depth, frequency, or reliance over time?
D6: Prompt Structure
Is the user's prompt discipline degrading over time?
Kit 2 · System → User
How the system talks to the user. Sycophancy patterns, warmth escalation, register drift, framing imposition, emotional initiation.
D1: Sycophancy Language
Is the system inflating praise, agreeing without basis, or burying dissent?
D2: Assumed Familiarity
Is the system performing prior knowledge of the user it doesn't have?
D3: Epistemic Overreach
Is the system claiming confidence beyond what its training supports?
D4: Autonomy Erosion
Is the system nudging the user toward its preferred direction rather than the user's?
D5: Register Drift
Is the system's register drifting from formal-analytical toward informal, affective, or relationally warm modes?
D6: Framing and Agenda
Does the system answer what was asked, or reshape the question and set the agenda?
D7: Emotional Initiation
Does the system respond to what the user brings, or open emotional doors the user left closed?
Kit 3 (System → Subject Matter) and Kit 4 (User → Subject Matter) are in development.