Coral CLI Skill

coral sql

Discovery-First Workflow

1. List available tables:

SELECT schema_name, table_name, description, required_filters, guide FROM coral.tables;

1. Read the

   guide

   column — it contains per-table query patterns and gotchas. Use it.
2. Check required filters for the table you want to query. Queries against tables with required filters will fail without the corresponding WHERE clauses.

SELECT column_name, data_type, is_required_filter, is_virtual, description
FROM coral.columns
WHERE schema_name = '<source>' AND table_name = '<table>'
ORDER BY ordinal_position;

1. Inspect source inputs when config affects the query or answer. This matters when you need source-specific values such as Datadog's

   DD_SITE

   or Sentry's

   SENTRY_ORG

   to build absolute links, explain account scope, or debug missing configuration.

SELECT key, kind, value, default_value, hint, required, is_set
FROM coral.inputs
WHERE schema_name = '<source>'
ORDER BY key;

1. Then query:

coral sql "SELECT <columns> FROM <source>.<table> WHERE <required_filters> LIMIT 10"

Query Guidance

• Virtual columns: Filter-only columns accepted in WHERE clauses but returning NULL in results. Check

  is_virtual

  in

  coral.columns

  .

• coral.inputs

  : Use it to inspect per-source variables and secrets before making assumptions about URLs, org names, regions, or other source config. Secret rows always return

  value = NULL

  ; use

  is_set

  to confirm whether a secret is configured.
• Cross-source joins: Standard SQL JOINs work across source schemas. Cross-source joins execute in memory after source scans complete.