The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/setup.sh script executes shell commands and Python code to programmatically modify AI agent configuration files. Specifically, it updates ~/.claude/settings.json, ~/.codex/config.toml, and ~/.gemini/settings.json to include a new MCP server entry pointing to http://localhost:8080/mcp.
[EXTERNAL_DOWNLOADS]: The skill instructs the user to install a Unity package from a remote repository (https://github.com/CoplayDev/unity-mcp.git). This source is external to the provided vendor and is not recognized as a trusted organization.
[INDIRECT_PROMPT_INJECTION]: The skill provides tools like read_console that ingest external data (Unity Editor console logs) into the agent's context. This creates a surface where malicious log content from a Unity project could influence the agent's behavior.
Ingestion points: read_console (reads runtime logs and error messages from the Unity Editor).
Boundary markers: None identified in the provided skill instructions.
Capability inventory: The skill has broad capabilities including Bash access, file Write operations, and script management tools (create_script, manage_script, script_apply_edits).
Sanitization: There is no explicit evidence of sanitization or filtering for the data read from the Unity console logs.
[DATA_EXPOSURE]: The skill accesses sensitive application configuration files located in the user's home directory (.claude/settings.json, .codex/config.toml, .gemini/settings.json) to perform its setup routine.

unity-mcp