# SCHEDULE_HELD

> the scheduler waits until a person has run this asset's code

**Warning** · category **coordination** (another process holds the database or the asset) · exit **0** (ok)

The scheduler waits until a person has run this asset's code.

The scheduler only runs code a person has run. An asset is held when it is new, or when its code changed since
it was last run by hand: its file, the lib/ code it imports, or its packages. The tick skips it and it stays
due; croft status, croft context and croft schedule status show it as held, with how long ago it was edited.
A file that no longer loads is held too, until it is fixed and run.

What to do: check the change, then run it by hand once.

- croft validate --json, then croft preview <asset>: read columns, checks, diff and samples. A preview that
  succeeds also releases the asset.
- croft run <asset>: a run from a terminal or from Claude Code releases it, and the scheduler runs it from its
  next fire on.

Ask the user first when the change is theirs to approve (a new source, a transform that pays per row). A
scheduled run never releases anything by itself.

The same page is in the CLI: `croft docs SCHEDULE_HELD`.

Source: https://croft.zabaca.com/docs/reference/errors/schedule-held/