Database ingestion config
On Entry
Tell the user:
Database ingestion config — this file is optional and it is read at conversion time. Skip it and relational objects keep their converted names. I'll write
source/_etl/database-ingestion-config.jsononly after you choose the rules and the validator accepts them.
Step 1: Ask
Ask via ask_user_question (multiSelect = false):
"Relational sources can be selected for Snowflake landing before conversion. What should I do with
database-ingestion-config.json?"
- Author — write a new schema 1.0 file.
- Edit — load the file that is already there.
- Skip — leave conversion unchanged.
Wait for the user's response — do not proceed until they choose.
Skip: leave the directory untouched. Tell the user conversion will not classify relational objects as ingestion, then return to ../SKILL.md Step 3.
Edit when source/_etl/database-ingestion-config.json is absent: say so and ask again with Author and Skip only.
The engine reads exactly one file of that name under source/_etl/. Write that path.
Step 2: Collect the document
Load references/schema.md and collect only the fields the user names. Leave out a field they did not ask for, including engine defaults.
Step 3: Validate, then write
From this skill's scripts/ directory, validate before the file is written:
python3 database_ingestion_config.py validate -
Pass the JSON on stdin. On failure the script prints one message that includes the JSON path. Show that message, correct the document, and validate again. Conversion stays at this step while the on-disk file is invalid.
On an edit, confirm the rewrite kept the user's fields:
python3 database_ingestion_config.py round-trip source/_etl/database-ingestion-config.json
Keys that were present stay present. Keys they did not set stay absent. Write the file only after validate exits 0.
Step 4: Return
Tell the user whether the file was written, edited, or skipped, and the path when one exists. Return to ../SKILL.md Step 3. Conversion is that skill's next step.