The Agent Skills Directory

REMOTE_CODE_EXECUTION (HIGH): The setup scripts dotnet-runtime.sh and dotnet-runtime.ps1 download and execute scripts from https://dot.net/v1/dotnet-install.sh and https://dot.net/v1/dotnet-install.ps1. These scripts are executed directly with bash or powershell to install the .NET SDK. This source is not on the provided list of trusted GitHub organizations.
DATA_EXFILTRATION (HIGH): The skill instructions in SKILL.md explicitly direct the agent to output the value of the SOLARWINDS_API_TOKEN environment variable (e.g., echo "$SOLARWINDS_API_TOKEN") during the installation check. This exposes sensitive credentials to the agent's context and conversation log.
COMMAND_EXECUTION (MEDIUM): The setup.sh and setup.ps1 scripts modify user configuration files (such as .bashrc, .zshrc, and PowerShell profiles) to persist PATH and environment variable changes. While this is standard installer behavior, it constitutes a persistence mechanism that modifies the user's environment.
EXTERNAL_DOWNLOADS (MEDIUM): The skill installs a .NET tool from a local package directory (./tools) and downloads the .NET installer from an external domain (dot.net).
INDIRECT_PROMPT_INJECTION (LOW): The skill ingests log data from the SolarWinds API. Malicious log content could potentially contain instructions intended to influence the agent's subsequent actions.
Ingestion points: Output of the logs command (SolarWinds API data).
Boundary markers: None present in the search tool output.
Capability inventory: The skill can execute arbitrary CLI commands via the setup scripts and the logs tool.
Sanitization: No explicit sanitization of log content is observed before presentation to the agent.

solarwinds-logs