The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains explicit instructions in SKILL.md directing the AI agent to "automatically" modify system configuration files when a user requests setup. This encourages the agent to bypass standard verification steps for sensitive file operations.
[COMMAND_EXECUTION]: The scripts/setup.sh file executes shell commands and inline Python scripts to modify AI agent configuration files, including ~/.claude/settings.json, ~/.codex/config.toml, and ~/.gemini/settings.json. Modifying these files is a sensitive operation as they define the agent's behavior and security boundaries.
[EXTERNAL_DOWNLOADS]: The skill references and encourages the installation of an external Unity package from https://github.com/CoplayDev/unity-mcp.git. While intended for functionality, users should verify the contents of third-party repositories before installation.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection.
Ingestion points: Data enters the agent context via read_console (Unity logs) and get_tests results.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present in the processing logic.
Capability inventory: The skill has access to high-privilege tools including Bash, execute_custom_tool, and manage_editor across its toolset.
Sanitization: No evidence of sanitization or filtering of the Unity console output or test data before it is processed by the agent was found.

unity-mcp