Agent Skills: One-Shot PySpark from the CLI

Agent workflows are not one-shot operations. The agent loops, retries, adjusts. Spark's cold start makes that loop impractical. A fast CLI changes what agents can actually attempt.

Each script bootstraps within a second? That's pretty cool!

The one-shot CLI is super useful on its own as well.

agent "build me a churn model from this data" = YES PLEASE