Captures reusable metrics, segments, and joins into ktx semantic layer YAML sources.
npx skills add https://github.com/Kaelio/ktx --skill sl-capture
Please help me install this Agent Skill. Skill: sl_capture Source: https://github.com/Kaelio/ktx/tree/main/packages/cli/src/skills/sl_capture Command: npx skills add https://github.com/Kaelio/ktx --skill sl-capture
Analytical agents repeatedly rediscover the same business metrics, join patterns, and derived dimensions in raw SQL. This Skill defines when and how to persist those reusable patterns into ktx's semantic layer so future queries reuse governed definitions instead of ad-hoc SQL.
sl_discover, sl_edit_source, and sl_write_source, with inherits_columns_from for slim standalone SQL sources.avg_order_value measure overlay on the orders source and validates it with sl_validate.Ask the agent to capture the reusable metric or join pattern from the current conversation into the semantic layer and validate the resulting source.
High-intent search queries and answers about installing and using this skill.
Run sl_discover on the target name first; if a manifest schema exists, write an overlay with sl_write_source containing only name and measures. For edits to existing sources, use sl_edit_source with exact-string yaml_edits, then run sl_validate.
Capture business metric aggregations, derived multi-signal calculations, reusable join patterns, and computed dimensions. Skip simple previews, trivial counts, one-off explorations, and measures that duplicate existing ones.
No. Time anchors and value lists belong in callers' query-time filters or in named segments, not inlined in measure expressions. Bake in a date only for regulatory cutovers, contractual boundaries, or one-time events.
Overlays patch manifest-backed tables with measures, segments, joins, and column_overrides while inheriting columns and grain. Standalone sql sources are only for metrics needing per-entity derivation or multi-step CTEs, and must target names not in the manifest.
Expression SQL must match the warehouse dialect and the column's manifest type; comparing TIMESTAMP columns to DATE-arithmetic results fails on BigQuery. The inline validator runs a LIMIT 1 probe per measure, so fix the expression and retry.
Use wiki-only edits when documenting a measure's business definition, owner, or policy without changing its SQL. Any request to change a measure's expression or filter requires an sl_edit_source or sl_write_source call.
AradotsoBuild a semantic context layer for AI agents querying data warehouses.
mirakuiDefine reusable business metrics in YAML for Databricks Runtime 17.2+.
nrakowDefine semantic models and metrics for dbt Semantic Layer, MetricFlow, and Cube.
LazyIsEfficientAnalyze YouTube channels to detect outlier videos and extract title packaging patterns.