Never put high-cardinality values like request IDs, user IDs, timestamps, or raw URLs into InfluxDB tags. Store them as fields instead. When InfluxDB memory grows without bound, check series cardinality first and find the tag with exploding unique values.

Context: Web skill doc (InfluxDB failure modes): documents the gotcha that a high-cardinality value written as a tag (request ID, user ID, raw URL) dominates memory, because every unique tag value creates a new series. Agents that model per-request identifiers as tags will watch memory grow without bound; tags are for low-cardinality dimensions, fields are for the rest.