Skip to content

Bulk Endorse or Deprecate Catalog Objects

Stewards often need to flag objects in bulk: “endorse all fields stewarded by this person”, “deprecate every order_dt column in the raw schema.” Doing it one object at a time in the UI doesn’t scale. This recipe builds an agent that finds the matching objects and applies trust flags through the Trust Check Flags API, with a human review step before anything is written. Deployed at two enterprises, with an approval-workflow variant in demand at several more.

Estimated cost0.25ACUper object

The catalog search is metered; the flag writes go through custom HTTP tools, which aren’t. A 500-object run pays for the search, not the writes. At 0.25 ACU per metered action; confirm actual spend on the Usage page.

  1. Create HTTP tools for the flag API

    Create custom HTTP tools for creating (and removing) trust flags via POST /integration/flag/. One tool per action keeps the agent’s choices simple.

  2. Build the agent

    Prompt pattern: find objects matching the user’s criteria, present the full list for confirmation, and only flag after the user approves. Add the search tool and your flag tools.

  3. Run it with review

    Ask: “Deprecate all order_dt fields in the raw schema.” Confirm the list, then let the agent apply the flags.

  • Custom HTTP tools send request bodies as JSON, but the M2M token endpoint requires a form-encoded body. Configure the tool with authentication None and pass a 24-hour API access token in a header instead.
  • Endorsements don’t accept comments, and flags can’t carry approval documentation — track the approval trail outside the flag (for example, in a custom field).
  • Approval-workflow variant: a first agent proposes flags into a review queue (custom field), a second applies approved ones on a schedule.