dotnet-msbuild-authoring

Pass

Audited by Gen Agent Trust Hub on Mar 5, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides numerous examples for using the MSBuild <Exec> task to run external shell commands, including dotnet tool run, sqlcmd, and signtool. These instructions teach the agent how to implement arbitrary code execution within a build pipeline.
  • [PROMPT_INJECTION]: The guidance on using item metadata (such as %(Identity)) within shell commands establishes an indirect injection surface. If the agent follows these patterns to build a project containing maliciously named files, it could inadvertently facilitate command injection.
  • Ingestion points: File system globs (e.g., migrations/*.sql) used to populate ItemGroup collections in SKILL.md.
  • Boundary markers: None; the examples do not suggest delimiters or instructions to ignore embedded commands in file metadata.
  • Capability inventory: Includes the <Exec> task for shell command execution and MSBuild property functions that allow calling .NET static methods (e.g., System.IO.Path, System.Environment, System.Text.RegularExpressions.Regex).
  • Sanitization: The skill does not provide guidance on escaping or validating file names/metadata before interpolating them into shell commands.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 5, 2026, 09:30 PM