The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes Power BI report metadata (visual.json, page.json, report.json) which are external data sources that could potentially harbor malicious instructions.
Ingestion points: The agent reads and analyzes PBIR JSON structure files as documented in references/pbir-structure.md.
Boundary markers: No explicit delimiters or instructions to ignore embedded content were found in the provided files.
Capability inventory: The skill assumes or suggests the use of powerful CLI tools such as pbir, fab (Fabric CLI), and te (Tabular Editor) for report conversion, model querying, and deployment (SKILL.md, important/MENTAL-MODEL.md).
Sanitization: There is no evidence of automated sanitization or filtering of external JSON content before it is processed by the agent.
[SAFE]: The skill references official Microsoft domains (e.g., developer.microsoft.com) and GitHub repositories for schema validation and documentation fetching, which are recognized as trusted technology services.
[SAFE]: Standard Base64 encoded image data is used within theme files (e.g., SqlbiDataGoblinTheme.json) for legitimate UI backgrounds and does not contain obfuscated code or commands.
[SAFE]: Example connection strings in definition.pbir utilize integrated security and do not expose sensitive hardcoded credentials.

pbir-format