Help: Need to demonstrate 'human-in-the-loop' for Claw in a regulated industry.

Hey everyone,

I’m working on a compliance project for a financial services client. We're implementing Claw (the data lineage tool) to help with audit trails for our data pipelines, which is a requirement for frameworks like SOC 2. Our auditors are asking us to specifically demonstrate a "human-in-the-loop" review for any automated lineage changes or anomaly detection.

From my reading, Claw can auto-discover lineage, but the guidance around formal human review processes seems a bit high-level. I’m trying to figure out the practical, day-to-day mechanics.

Here’s my current thinking and where I'm stuck. We have a scheduled Claw job that runs weekly. I can get a report of new or modified lineage edges, but I need to:
1. Flag these changes for a human reviewer (like a senior analyst or data steward).
2. Have them officially "approve" or "reject" the change in a way that creates an audit log.
3. Tie that log back to our GRC platform.

Is anyone doing this? My initial idea is to use the Claw API to fetch changes and then build a simple review workflow. Maybe something like this:

```sql
-- Example: Query to get recent lineage changes not yet reviewed
SELECT change_timestamp,
source_object,
target_object,
change_type
FROM claw.operational.lineage_changes
WHERE review_status IS NULL
AND change_timestamp > CURRENT_DATE - INTERVAL '7 days';
```

But then I need to build the review interface and logging part. Should I be looking at integrating with something like Jira Service Desk, or is there a simpler way to generate a report with a manual sign-off sheet? I'm worried about creating extra manual work for the team.

Also, for those in regulated industries: does the review need to happen for *every* auto-discovered change, or is a risk-based sampling approach acceptable to auditors? Any examples of how you structured this would be super helpful.