# GTM workflow change list: <workflow name>

Configuration changes to vendor tools (tables, columns, workflows, campaigns, assignment rules, policies). Record-level writes (updating contacts, creating companies) use agent-write-safety's change proposal with the `check_writeback.py` change list attached.

- Baseline: <systems, settings read, read time>
- Review or spec: <path>
- Prepared by: <agent run or person>. Applies: <name>. Verifies: <name>
- Status: draft, not applied. Nothing in this list has been changed in any system.

## Summary by consequence

- Contact without permission (suppressed, customer, open opportunity, unapproved region): <CR numbers>
- Data overwritten, cleared or duplicated: <CR numbers>
- Money spent without an accepted result: <CR numbers>
- Missing work (unrouted items, lost requests): <CR numbers>
- Hygiene (separate release): <CR numbers>

## CR-<n> <short title> (<rule ID> <FAIL | WARN>)

- System and object: <tool>, <table and column | workflow | campaign | assignment rule | policy file>, ID <placeholder or internal ID>
- Current setting: <quote it from the export or screenshot>
- Change: <exact edit, field by field>
- Why: <the defect, with the rule ID and the evidence line>
- Could break: <what depends on the current setting, or "nothing that is correct today">
- Depends on: <other CR numbers, or none>
- Test: <steps, expected result, acceptance case IDs from acceptance-cases.md>
- Test approval and data: <who approved, test records, controlled inboxes, sandbox or test portal>
- Records the test creates: <list, and who removes or excludes them>
- Rollback: <previous setting or version to restore>
- Rollback does not undo: <messages sent, credits spent, records created or changed, tasks assigned>
- Owner sign-off: <name, date>

## Not changed, needs a decision

- <finding>: <the question, and the owner who answers it>

## Release evidence to ask for

The person who applies each item closes it with three answers:

- Did the setting change as written? (screenshot or export after the change)
- Did runtime behaviour match on test records? (acceptance case results)
- Is the destination state correct on read-back? (CRM records, sequencer leads, usage figures)

A yes to the first question does not stand in for the other two.