The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the Azure CLI (az rest) and specialized Fabric tools (fab, powerbi-report-author) to manage semantic model items, trigger refreshes, and discover lakehouse metadata. These tools are used correctly within the scope of their intended administrative and development purposes.
[EXTERNAL_DOWNLOADS]: Deployment and CI/CD references include instructions to install standard developer utilities such as ajv-cli via npm and TabularEditor.TOM via the .NET tool registry. These downloads originate from well-known registries and target established community tools for Power BI development.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests schema metadata and data previews from external databases and lakehouses to inform its modeling decisions.
Ingestion points: SKILL.md identifies lakehouse table exploration and data preview operations as primary data inputs.
Boundary markers: The skill provides rigorous 'Must/Avoid' guidelines and a 'Validation Checklist' to ensure model integrity and correctness.
Capability inventory: The skill is capable of performing network operations and updating model definitions via az rest, as well as executing DAX queries against connected models.
Sanitization: The instructions emphasize strict adherence to star schema principles and naming conventions to maintain data consistency.
[SAFE]: The skill demonstrates best practices for secret management by instructing users to use environment variables and GitHub Action secrets rather than hardcoding credentials. The use of Base64 encoding in the API reference is a functional requirement for data transport and does not constitute malicious obfuscation.

power-bi-semantic-model