You want a workflow to audit an entire repository and you cannot tell in advance what that will cost. Someone suggests writing a size guideline into the script and letting it run overnight in the background. What is the sensible first move?

The principle — the way to price an unknown run is to buy a small piece of it, because nothing in the arrangement will stop a large one for you.

Why the key is correct — the advice is explicit: to gauge the spend before committing to a large task, run the workflow on a small slice first — one directory instead of the whole repo, or a narrow question instead of a broad one. That works because a workflow is worth reaching for when a task needs more agents than one conversation can coordinate, or when you want the orchestration codified as a script you can read and rerun — the same script, pointed at less. Two facts close the alternatives. Runs count toward your plan's usage and rate limits like any other session, so the background is not a cheaper place to run. And the guideline in the script is sent to Claude as advice, not a cap, so it shapes how the work is divided without bounding what it costs.

Why the others are wrong — a guideline is not enforced, so it is not a budget. Background execution does not move a run outside your limits. And the spend is not unknowable — one directory answers it.

Remember this — price a big run by running a small one. A guideline is advice; the bill is yours either way.

Sources — Anthropic's Claude Code documentation on orchestrating subagents at scale with dynamic workflows.