The Agent Skills Directory

[Command Execution]: The skill utilizes the npx fabric-app-data command to execute DAX queries against semantic models.
Evidence: Found in SKILL.md under the "Running Discovery Queries" section: npx fabric-app-data query <alias> --query '<DAX>'.
Context: This command is used for its intended purpose of metadata discovery within the Microsoft Fabric environment and is consistent with the skill's stated functionality.
[Indirect Prompt Injection Considerations]: The skill is designed to ingest and process metadata (table names, column names, measures, etc.) from external Power BI semantic models.
Ingestion points: Metadata is retrieved via INFO.VIEW.* and INFO.* functions in references/discovery-queries.md.
Boundary markers: The skill recommends using SELECTCOLUMNS and FILTER to narrow results, which acts as a form of logical data delimitation.
Capability inventory: The skill's primary capability is executing read-only DAX queries to retrieve schema information.
Sanitization: While the skill doesn't explicitly mention string sanitization for retrieved metadata, the nature of the data (database schemas) and the read-only execution environment significantly limit potential risks associated with malicious data injection.
Context: Any tool that reads data from external, potentially untrusted sources (like a semantic model schema that could be modified by a third party) has an inherent exposure to indirect prompt injection. The skill mitigates this by focusing on structured metadata discovery rather than processing arbitrary text.

schema-discovery