zone-of-indifference-override

Preventing automatic AI compliance requires surfacing contextual mismatch and intent ambiguity before execution. This skill implements a structured override format that pauses Claude to verify ambiguous, high-stakes requests instead of executing them blindly.

A zone of indifference override detects when a request feels off despite being technically safe and surfaces concerns before execution. It uses contextual mismatch and downstream harm signals to prompt focused confirmation without refusing legitimate work.

Verifying ambiguous intent involves implementing a non-blocking escalation surface that shows the request, concern, risk, and question. This guides intent verification to ensure controlled action in high-stakes or ambiguous delegation scenarios.

Contextual safety checks work in delegation chains to catch edge cases where safe requests could cause downstream harm. The override surface integrates with related safety controls like cognitive-friction and reversibility-gate to ensure controlled action.

Intent clarification is needed when contextual mismatch or intent ambiguity signals appear in Claude's responses. It prevents automatic compliance in high-stakes requests by surfacing concerns and prompting focused confirmation before execution.

The override surface format uses non-blocking escalation to surface concerns without refusing legitimate work. It guides intent verification while allowing safe tasks to proceed, ensuring controlled action only when contextual mismatch or downstream harm signals are detected.