The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents patterns for ingesting data from external sources, including PostgreSQL databases and streaming data pipelines.
Ingestion points: etlPipeline (extracts data from Postgres), streamInserts (processes an external dataSource), and the CDC listener (receives market_updates notifications).
Boundary markers: None identified in the provided templates.
Capability inventory: The skill uses clickhouse.query and clickhouse.insert to execute SQL and batch operations.
Sanitization: The provided snippets show basic transformation logic (mapping values) but do not include explicit sanitization of input data before database insertion. This is standard for educational templates focused on architecture rather than validation logic.
[SAFE]: The code snippets follow security best practices by utilizing environment variables (process.env.CLICKHOUSE_PASSWORD, process.env.DATABASE_URL) to handle credentials, avoiding the risk of hardcoded secrets.

clickhouse-io