The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to use the wmill CLI for various operations, including wmill script preview, wmill script run, and wmill generate-metadata. These commands allow the agent to manage local files and interact with the Windmill platform as part of a standard development workflow.\n- [DYNAMIC_EXECUTION]: The agent is instructed to use wmill script preview to execute scripts locally. This enables the agent to test BigQuery queries and script logic before deployment, which is a core function of the skill.\n- [INDIRECT_PROMPT_INJECTION]: The skill identifies that script schemas may contain a prompt_for_ai field. The agent is explicitly told to follow these instructions when choosing inputs, which represents an indirect injection vector from external metadata.\n
Ingestion points: The prompt_for_ai string located in .script.yaml files.\n
Boundary markers: No specific delimiters or safety warnings for handling these embedded instructions are provided.\n
Capability inventory: The agent can execute local code via wmill, modify local filesystem metadata, and perform remote deployments via wmill sync push.\n
Sanitization: The skill does not specify any sanitization or validation protocols for the content of the prompt_for_ai field.

write-script-bigquery