cost-estimationlisted

# Cost Estimation Estimate, *before* work runs, the token usage and time a task will consume flowing through the orchestrator pipeline — and a dollar cost **only when observed actuals exist to ground it**. This is the **prospective** counterpart to the **retrospective** `cost-tracking` skill. `cost-tracking` *records* what was spent into `observability/costs/YYYY-MM-DD-costs.md`; this skill *reads* those same snapshots as its $/token ground and *predicts* what a task will spend. The two are siblings — a reader who finds one should find the other (see [Sibling relationship](#sibling-relationship-to-cost-tracking)). The day-one deliverable is a **token + time estimate**. Both axes are fully groundable from `MODEL_ROUTING.md` today, so they are the honest first-version value of this capability. The dollar figure is an **actuals-gated enhancement** — present and grounded only when an `observability/costs/` snapshot supplies a usable $/token rate, otherwise omitted with an explicit disclosure (never a forced list-price guess). This skill is **methodology and a format contract**. It describes how an estimate is derived and what an estimate record must contain. It does not dispatch an agent, write a file, or decide go/no-go — see [What this skill does NOT do](#what-this-skill-does-not-do). ## An Estimate Is a Range, Never a Point The material decision this skill locks: **an estimate is a range with disclosed confidence, never a point value.** A point value invites false preci