Write and understand SQL transforms for Turbo pipelines.
SkillScore95Maintained
Automatic
◆Data Engineer pack
Last verified 8/6/2026
What it changes
Passed 20 of 21 test cases with the skill, and 6 of 21 without it.
Without skill27%
With skill95%
Checks it passed that the model alone missed
The response configures batch_size in the sink YAML block.
The response uses to_i256 to cast signed numerical values.
The response uses dynamic_table_check in the SQL WHERE clause.
and 27 more in the full report
16 fixed5 passed either way0 still failing1 broke
Pass-rate lift
+68 pp
Tokens
+272%
Agent turns
0%
Response time
-55%
Model gemini-3.6-flashMeasured 8/6/2026v121 of 22 cases passed with skill · 21 comparable
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 15 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 15
Prompt
I am transforming EVM log records into a downstream PostgreSQL sink, but my update and delete operations in the database are overwriting rows incorrectly. Which system metadata column must be preserved in SQL transforms to maintain upsert semantics?
What was checked, without and with the skill
Checked
Without skill
With skill
The response includes the _gs_op column in the transform SELECT statement.
Missed
Passed
case-09 · gemini-3.6-flash · 8/6/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md780 lines2 files
Turbo Transforms
Write, understand, and debug SQL transforms for Turbo pipeline configurations.
Triggers
Invoke this skill when the user:
Asks "how do transforms work" or "what can I do with transforms"
Wants to decode raw EVM logs or contract events
Needs to filter, reshape, or combine blockchain data
Asks about evm_log_decode, _gs_log_decode, U256/I256 math, or other Goldsky SQL functions
Wants to chain multiple transforms together
Needs help writing SQL for a pipeline YAML
Is debugging a transform that isn't producing expected results
Wants to combine multiple event types into a single output
Mentions /turbo-transforms
Agent Instructions
Step 1: Understand What the User Needs
Determine the transform goal:
Decode raw logs → Guide them through ABI-based log decoding with evm_log_decode()
Filter data → Help write WHERE clauses (address, event type, block range, etc.)
Reshape columns → Help with SELECT, CAST, CASE WHEN, string manipulation
Combine data → Help with UNION ALL across multiple event types or sources
Chain transforms → Help build multi-step processing pipelines
Understand existing transforms → Walk through the SQL and explain each part
Step 2: Write or Explain the Transform
Use the patterns and references below to help the user. Always ensure:
Every transform has type: sql and primary_key: id (or appropriate column)
SQL references source or transform names in the FROM clause
Column names match the source schema
CAST operations use correct target types (DOUBLE, VARCHAR, etc.)