Imported from snowflake-labs/subagent-cortex-code (
integrations/cursor/SKILL.md) via skills.sh. Install upstream withnpx skills add snowflake-labs/subagent-cortex-code --skill cursor. Copyright stays with the author (Proprietary. See LICENSE for complete te).
Cortex Code Integration
Routes Snowflake queries to Cortex Code agent with conversation context enrichment.
When to Use
Use this skill when the user asks about:
- Snowflake databases, warehouses, schemas, tables, views
- SQL queries for Snowflake data ("How many databases?", "Show top customers")
- Data quality checks, validation, profiling
- Cortex AI features (Cortex Search, Cortex Analyst, ML functions)
- Semantic views, data modeling
- Snowpark, dynamic tables, streams, tasks
- Snowflake security, roles, policies, governance
Instructions
Step 1: Build Enriched Context
Before executing, build an enriched prompt that includes:
-
Conversation Context (if relevant):
- Previous 2-3 exchanges from this conversation
- Any Snowflake-specific details already discussed
- User's stated goals or requirements
-
User's Question:
- The current query being asked
Step 2: Execute with Context
Pass the enriched context to Cortex Code:
python3 scripts/execute_cortex.py \
--prompt "# Conversation Context
[relevant prior exchanges if any]
# Current Question
[USER'S QUESTION]" \
--envelope "RO"
Example with Context
If user previously asked "Which databases have stock data?" and now asks "Show me the schema for the main table":
python3 scripts/execute_cortex.py \
--prompt "# Recent Context
User previously identified databases with stock data: DB_STOCK, FINANCE__ECONOMICS
# Current Question
Show me the schema for the main table in DB_STOCK" \
--envelope "RO"
Example without Context
For standalone questions:
python3 scripts/execute_cortex.py \
--prompt "How many databases do I have in Snowflake?" \
--envelope "RO"
How It Works
- Agent builds enriched prompt with conversation context
- Routes to Cortex Code agent (headless execution)
- Cortex executes
cortexCLI with stream-json format - Runs SQL queries via
snowflake_sql_executetool - Returns formatted results with analysis
Configuration
- Approval mode: prompt by default; request user approval before execution
- Security envelope: RO for reads; RW only for approved writes
- Connection: Uses default Snowflake connection from cortex CLI
