dotnet-inspect-package-skills
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on the execution of shell commands, specifically using a tool called
dnx dotnet-inspectto interact with the .NET ecosystem and project files.\n- [INDIRECT_PROMPT_INJECTION]: The primary function of this skill is to ingest and apply instructions (agent skills) found in external NuGet packages and project assets. This creates an attack surface where a compromised or malicious package could inject instructions to override agent behavior.\n - Ingestion points: External NuGet packages (e.g.,
Markout@0.35.2) and local project metadata files (project.assets.json).\n - Boundary markers: The skill documentation notes that the
dotnet-inspecttool implements validation logic and a 'containment' system to filter or omit potentially unsafe content, indicated by the marker[Text omitted: required containment].\n - Capability inventory: The skill utilizes shell command execution, file system reads, and the ability to write or overwrite files in the repository (e.g., via redirection
> skills/...or the--outputflag).\n - Sanitization: The skill relies on the underlying
dotnet-inspecttool to validate metadata (names and descriptions) and enforce content safety policies before the agent processes the external skills.\n- [EXTERNAL_DOWNLOADS]: The skill facilitates the discovery and fetching of packages from the NuGet registry. While NuGet is a standard developer service, the runtime acquisition of third-party content is an external dependency pattern that should be monitored.
Audit Metadata