The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests event data from external sources, which is a surface for indirect prompt injection.
Ingestion points: Data is queried from the PostHog events table and dbt staging models.
Boundary markers: The instructions explicitly mandate treating candidate event names as quoted data to prevent them from being interpreted as instructions.
Capability inventory: The skill uses tools for view creation and SQL execution.
Sanitization: The skill recommends treating event names as untrusted input and requires user confirmation before persistent view creation.

modeling-activation-metrics