The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted event names and properties from external analytics sources to build conversion models.
Ingestion points: Data enters the context via the events table in references/posthog/funnel_conversion.sql and staging references in references/dbt/stg_funnel_events.sql.
Boundary markers: The skill explicitly warns to treat event names as quoted data and not as instructions in SKILL.md.
Capability inventory: The skill's capabilities are restricted to generating and executing SQL queries for data warehouse views; no subprocess execution, network operations, or file system writes are present.
Sanitization: The instructions recommend verifying event names using read-data-schema and obtaining user confirmation before creating persistent views.

modeling-conversion-metrics