| name | lineage-diff |
| description | Compare column-level lineage between two versions of a SQL query to show added, removed, and changed data flow edges. |
Lineage Diff
Requirements
Agent: any (read-only analysis)
Tools used: lineage_check, read, bash (for git operations), glob
Compare column-level lineage between two versions of a SQL model to identify changes in data flow.
Workflow
-
Get the original SQL — Either:
- Read the file from disk (current committed version)
- Use
git show HEAD:path/to/file.sql via bash to get the last committed version
- Accept the "before" SQL directly from the user
-
Get the modified SQL — Either:
- Read the current (modified) file from disk
- Accept the "after" SQL directly from the user
-
Run lineage on both versions:
- Call
lineage_check with the original SQL
- Call
lineage_check with the modified SQL
-
Compute the diff:
- Added edges: Edges in the new lineage that don't exist in the old
- Removed edges: Edges in the old lineage that don't exist in the new
- Unchanged edges: Edges present in both
-
Report the diff in a clear format:
Lineage Diff: model_name
═══════════════════════════
+ ADDED (new data flow):
+ source_table.new_column → target_table.output_column
- REMOVED (broken data flow):
- source_table.old_column → target_table.output_column
UNCHANGED: 5 edges
Impact: 1 new edge, 1 removed edge
Usage
The user invokes this skill with a file path:
/lineage-diff models/marts/dim_customers.sql — Compare current file against last git commit/lineage-diff — Compare staged changes in the current file
Edge Matching
Two edges are considered the same if all four fields match:
source_table + source_column + target_table + target_column
The transform field is informational and not used for matching.
Use the tools: lineage_check, read, bash (for git operations), glob.
