Clarify Cost Appetite

Why this file exists

azure-to-aws had a first-class, explicitly-named “cost optimization appetite” question (its Q-A4 — Conservative / Balanced / Aggressive) feeding rightsizing-thresholds.json. gcp-to-aws had no equivalent named question — its cost posture was implicit, spread across the opt-in Category E (“Migration Posture,” disabled by default) right-sizing toggle. Promoting Azure’s version into this shared file, and wiring GCP’s Category E to read from it, gives both skills one explicit, always-visible cost-appetite knob instead of GCP’s silently-disabled-by-default right-sizing toggle.

The question

How aggressively should we right-size your AWS design versus your current capacity?

[A] Conservative — like-for-like capacity, lowest risk [B] Balanced — right-size where measured data supports it (default) [C] Aggressive — smallest defensible footprint

Consequence line: Balanced right-sizes only where measurement supports it. Aggressive can cut the estimate materially and needs load testing before cutover.

Interpret

Default: balanced.

With no utilization data, this answer changes nothing — say so on the sheet row rather than implying a right-sizing pass will run when there is no measured baseline to right-size from. This is the single biggest trap with this knob: an “Aggressive” answer over an inventory with no billing/utilization export produces the same design as “Balanced,” and the sheet must say that plainly rather than letting the user believe they bought a smaller estimate.

What a consuming skill supplies

1. Where the knob feeds. Azure feeds knowledge/estimate/rightsizing-thresholds.json directly. GCP feeds its Category E opt-in (ha_upgrade, right_sizing) — GCP’s Category E questions (Q-E1/Q-E2) become redundant with, and should be superseded by, this single knob: a GCP run that answers aggressive here should behave as if the user opted into Category E’s right_sizing: true, without asking Q-E2 separately. GCP’s Q-E1 (HA upgrade opt-in) is a distinct decision — this question governs sizing/instance-class aggressiveness, not availability tier — and stays a separate opt-in.
2. Disposition. PROPOSED in both skills; never ESSENTIAL — a sensible default (Balanced) always exists and the consequence of guessing wrong is a report caveat, not a wrong architecture.