package-add

Warn

Audited by Gen Agent Trust Hub on Aug 21, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The package-add tool explicitly supports installing Unity packages from arbitrary Git URLs and local paths. Because Unity packages can contain scripts and assembly files that are compiled and executed by the Unity Editor upon installation, this provides a vector for remote code execution from untrusted sources.
  • [COMMAND_EXECUTION]: The skill facilitates the execution of shell commands through unity-mcp-cli to modify the project's manifest.json and trigger package resolution. It also suggests installing the CLI tool globally using npm install -g, which typically requires administrative privileges on many systems.
  • [EXTERNAL_DOWNLOADS]: The skill references the external tool unity-mcp-cli and instructs the agent or user to download it from the NPM registry. Furthermore, the tool's primary purpose is to download and integrate external code repositories into a project.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it accepts a packageId string without input validation, sanitization, or boundary markers. An attacker could embed a malicious Git URL in data the agent processes (such as a project requirement or a README file) to trick the agent into installing a compromised package.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 21, 2026, 12:15 PM