Skills Reference

Complete reference for all 23 SignalPilot plugin skills plus the 2 verifier agents. Each skill is a markdown knowledge file whose frontmatter description is its load trigger. Skills run on both the Claude Code plugin and the Codex plugin.

Skills

Skill	When it loads	What it covers
signalpilot	FIRST tool call when a message mentions dbt, SQL, database, or data pipeline (blocking requirement).	The SignalPilot MCP tools, the available skills, and the governed workflow for dbt projects, SQL queries, schema discovery, and database access.
dbt-workflow	FIRST, before any dbt project work.	The full 8-step dbt workflow: project scanning, skill loading, validation, macro discovery, research, technical spec, SQL writing, verification. Plus output-shape inference, incremental model handling, and what to trust in YML.
dbt-write	Step 2 of the workflow (always).	Column naming, type preservation, JOIN defaults, lookup joins, sibling models, materialization, packages, and filtering rules.
dbt-debugging	When dbt run or dbt parse fails.	YML duplicate patches, ref errors, passthrough model warnings, current_date fixes, DuckDB error messages, and zero-row diagnosis.
dbt-testing	When the task mentions tests/unit tests, or the scan finds unit_tests: in YML.	unit_tests YAML format, given/expect blocks, edge-case coverage, and the difference between unit tests and schema tests.
dbt-snapshots	When the task involves snapshots, SCD Type 2, or change tracking, or a snapshots/ dir exists.	Strategy selection, column casing, verification, and common pitfalls.
dbt-versioning	When the task involves versioning / v2 / backward-compatible changes, or YML has versions:.	versions YAML config, defined_in, latest_version, and ref() with version pins.
dbt-knowledgebase	When populating the knowledge base from dbt project research.	Proposes entries across all 6 categories at org, project, and connection scopes.
knowledge-base	Step 6 of the workflow.	Writing the per-task technical_spec.md: distills research into structured decisions. Retries read the existing spec instead of re-researching.
duckdb-sql	When hitting DuckDB syntax errors or writing DuckDB SQL.	DuckDB gotchas that differ from PostgreSQL/MySQL.
snowflake-sql	When writing Snowflake SQL or hitting Snowflake errors.	QUALIFY, LATERAL FLATTEN, semi-structured VARIANT, ILIKE, date functions, and time travel.
bigquery-sql	When writing BigQuery SQL.	UNNEST, STRUCT, ARRAY_AGG, DATE_DIFF/DATE_ADD, backtick-quoted refs, EXCEPT/REPLACE in SELECT, approximate aggregation, partitioned/wildcard tables.
sqlite-sql	When writing SQLite SQL.	substr/instr, `
domain-ecommerce	Step 2, for orders/products/discounts/returns/charges/spend tasks.	Transaction lifecycle, driving tables, status filtering.
domain-financial	Step 2, for revenue/invoices/ledgers/fiscal tasks.	Grain consistency, balance sheets, double-entry ledgers, fiscal-year boundaries, period-over-period calculations.
domain-healthcare	Step 2, for clinical/patients/encounters/diagnoses/costs tasks.	Encounter-based grain, clinical coding hierarchies, cost allocation, NULL semantics in clinical data.
domain-hr	Step 2, for employees/hiring/issues/SCD/tickets tasks.	SCD current-record filtering, issue-resolution metrics.
domain-marketing	Step 2, for campaigns/clicks/email/SMS/attribution tasks.	Attribution models, engagement funnel order.
domain-media	Step 2, for movies/sports/credits/rankings/content tasks.	Content catalogs, participation tables, ranking determinism.
domain-product	Step 2, for events/sessions/features/guides/analytics tasks.	Calendar spine cross-joins, date boundary caps, event-type pivoting, first-run NULL behavior.
sql-workflow	Before writing any standalone (non-dbt) SQL query.	Output-shape inference, efficient schema exploration, iterative CTE building, a structured verification loop (row count, NULL audit, fan-out, sample inspection), error recovery, saving to result.sql/result.csv, turn budget, and common benchmark traps.
write-report	Only when explicitly requested (model invocation disabled).	Generates an HTML report of dbt project work: decisions, SQL, queries, verifier results, and visual charts.

Agents

Both agents are dispatched in parallel in Step 8 of dbt-workflow. Both are strictly read-only: they return reports and fix nothing.

Agent	Role	Checks
verifier	Structure verification (read-only).	CHECK 1 table existence; CHECK 2 column completeness via map-columns + check_model_schema; CHECK 3 row count, fan-out, and cardinality via audit_model_sources; CHECK 4 non-deterministic SQL (ORDER BY NULL, ROW_NUMBER/RANK without ORDER BY); CHECK 5 source-table preservation for modified models.
value-verifier	Aggregate value verification (read-only).	CHECK 1 sample value spot-check vs siblings; CHECK 2 aggregate cross-validation via verify_model_values (COUNT(*) vs COUNT(DISTINCT) aligned to the column name); CHECK 3 status-column filtering (returns/cancellations/refunds excluded per the domain skill). Prescribes exact CHANGE: fixes on FAIL but never edits files.