The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill pulls the official metabase/metabase-enterprise Docker image from Docker Hub. This is a trusted resource provided by the vendor 'metabase'.
[COMMAND_EXECUTION]: The skill uses the Bash tool to execute docker pull and docker run commands. These are necessary to invoke the semantic checker and are restricted to the local workspace mount.
[INDIRECT_PROMPT_INJECTION]: The skill processes project YAML files and a metadata JSON file. While these are untrusted inputs, they are analyzed by a specialized Metabase binary rather than directly influencing agent logic in a dangerous way.
Ingestion points: Repository YAML files and .metadata/table_metadata.json.
Boundary markers: None explicitly defined, but the data is passed to a structured validator.
Capability inventory: Bash, Read, Glob, Grep are used to locate and process files.
Sanitization: Relies on the java -jar metabase.jar --mode checker command to validate the integrity of the data.

metabase-semantic-checker