How FLUID, STANDARD, STANDARD+, and LOCKDOWN modes control Copilot's response format, questioning, and validation.
Strictness modes are automatically set by the Risk Engine based on the computed risk level. They control three things: how much Copilot questions before acting, the output format schema, and which validation checklists run.
FLUID
Auto at L0Gate: OffMinimal overhead. Direct answers, no governance banners. Ideal for questions, explanations, and trivial code snippets.
Example output:
User: 'What is the difference between let and const?' Copilot: Concise direct answer. No headers, no risk classification.
STANDARD
Auto at L1Gate: If Axis D ≥ 2 → 1 questionOptional one-line plan summary before code. Compact agent notice. Triggers a single clarifying question only when significant ambiguity is detected.
Example output:
User: 'Add a loading spinner to the button' Copilot: [Optional one-line intent summary] [Implementation]
STANDARD+
Auto at L2Gate: Max 2 focused questionsFull structured 6-field output: Risk Classification, Intent Summary, Impact Analysis, Assumptions, Proposed Solution, Risk Mitigation, and a Validation Checklist.
Example output:
Risk Classification: L2 — Elevated | Score: 8/12 | Mode: STANDARD+ Intent Summary: ... Impact Analysis: ... Assumptions: ... Proposed Solution: ... Risk Mitigation: ... Validation Checklist: ...
LOCKDOWN
Auto at L3Gate: Full gate + confirm loopMaximum safety. 10-field deterministic output including Failure Modes, Required Safeguards, Implementation Plan, Rollback Plan, and Final Safety Confirmation (PROCEED / HOLD).
Example output:
Risk Classification: L3 — Critical | Score: 11/12 (A:3 B:3 C:2 D:3) | Mode: LOCKDOWN ... Final Safety Confirmation: HOLD — Migration is irreversible. Provide a rollback script before proceeding.
Override the auto-set mode at any point in a conversation:
"go fluid" → FLUID mode (no governance headers)
"standard+" → STANDARD+ (structured output)
"lockdown" → LOCKDOWN (maximum safety, full 10-field output)Safety is absolute