Bulk edits

Changing more than one dataset at a time is always preview → confirm → apply. That includes the in-app assistant and any script you write. Zero silent writes.

Preview

A preview states, in the same breath as the ask:

• Which project (and team)
• How many datasets
• Names (or a prefix plus a count if the list is long)
• The exact change (instrument=182, tag needs-review, “remove parameter sn”)

Example:

Add instrument=182 to 23 datasets in cable-qual with lot=15: … Proceed?

Apply

After an explicit confirm:

• The change is applied.
• The result reports updated, skipped (already had the value, missing permission, missing the parameter to delete), and failed (name + reason).
• Failures are not silently retried in a loop.

Viewers cannot apply. Members can, on datasets they can edit.

Typical operations

Ask	Preview	Apply
Add instrument=182 to lot=15	Filter, list, count	Set the parameter
Tag this notebook’s datasets needs-review	List those datasets	Get-or-create the team tag, apply
Remove sn from one dataset	Name + current value	Delete that parameter
Which of these are missing sn?	Read-only list — no confirm needed	—

Reads do not need a confirm. Mixed “tag these and tell me who lacks serial” is a read plus a write: preview the write, then apply, then answer the missing-serial list from the same filter.

Tags are team-wide

Get-or-create the tag by name on the team; do not create needs-review a second time because you are in another project. See Parameters and tags.