c-sharp-scripting

Pass

Audited by Gen Agent Trust Hub on May 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides instructions and examples for executing arbitrary C# scripts and Tabular Model Scripting Language (TMSL) commands through helper methods such as ExecuteCommand, EvaluateDax, and CustomAction. This functionality is the primary purpose of the skill for advanced model management.
  • [DATA_EXFILTRATION]: In examples/tables/format-power-query.csx, the script uses System.Net.Http to send Power Query (M) expressions to a third-party API (https://m-formatter.azurewebsites.net/api/v2) for code formatting. This behavior involves transmitting model metadata to an external, non-whitelisted service.
  • [INDIRECT_PROMPT_INJECTION]: The skill exhibits a vulnerability surface where it ingests potentially untrusted data from model metadata or local files and possesses powerful execution capabilities.
  • Ingestion points: ReadFile (various examples), EvaluateDax results (examples/evaluate-dax/execute-table-query.csx), and Selected object metadata (examples/measures/create_sum_measures.csx).
  • Boundary markers: Absent from prompts that interpolate object names or expressions.
  • Capability inventory: SaveFile, ExecuteCommand, EvaluateDax, CustomAction, and ExecuteReader across multiple scripts.
  • Sanitization: No explicit escaping or validation of ingested metadata before its use in further commands was observed.
Audit Metadata
Risk Level
SAFE
Analyzed
May 16, 2026, 12:57 PM
Security Audit — agent-trust-hub — c-sharp-scripting