c-sharp-scripting
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function involves analyzing and bulk-modifying Power BI model metadata, which serves as a potential vector for indirect prompt injection if the model contains malicious content.
- Ingestion points: The agent reads and processes model definitions (BIM and TMDL files) and C# scripts (referenced in SKILL.md and multiple example files).
- Boundary markers: No explicit markers or instructions are used to distinguish untrusted model metadata from trusted instructions during analysis.
- Capability inventory: The skill utilizes shell command execution (te CLI), file system read/write access (SaveFile/ReadFile), and network operations via standard .NET libraries within C# scripts.
- Sanitization: Model metadata such as descriptions and expressions are processed and written back to the model without sanitization (e.g., examples/bulk-operations/update_descriptions_from_comments.csx).
- [DATA_EXFILTRATION]: The script examples/tables/format-power-query.csx sends model M expressions (Power Query code) to an external third-party API (m-formatter.azurewebsites.net) for formatting. This transmits model configuration details, potentially including server names or infrastructure parameters, to an external domain.
- [DYNAMIC_EXECUTION]: The skill is designed to generate and execute C# scripts (.csx) at runtime using the Tabular Editor environment. This behavior is documented in SKILL.md and implemented across various examples for model manipulation.
- [COMMAND_EXECUTION]: The skill frequently invokes shell commands via the Tabular Editor CLI (te.exe) to perform model operations, as described in the execution methods of SKILL.md.
Audit Metadata