The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The installation process requires cloning a plugin from a third-party GitHub repository (github.com/youichi-uda/godot-mcp-pro) and acquiring a server package from external platforms including itch.io and Buy Me a Coffee.
[COMMAND_EXECUTION]: Users are instructed to execute manual shell commands, such as git clone, npm install, and npm run build, to set up the environment using code from unverified external sources.
[DYNAMIC_EXECUTION]: The skill defines tools like execute_editor_script and execute_game_script which allow the AI agent to generate and run arbitrary GDScript code at runtime within the Godot Editor or game instance.
[INDIRECT_PROMPT_INJECTION]: The skill exposes the AI agent to untrusted project data, creating a risk of indirect prompt injection.
Ingestion points: Data is ingested into the agent context through tools like read_script, get_output_log, get_filesystem_tree, and search_in_files found in SKILL.md.
Boundary markers: There are no explicit markers or instructions to delimit or sanitize data read from project files to prevent the execution of embedded malicious instructions.
Capability inventory: The agent possesses significant capabilities including file system modifications (create_script, edit_script, save_scene) and high-privilege code execution (execute_editor_script).
Sanitization: No sanitization or validation mechanisms are described for the content ingested from the local development environment.

godot-mcp-pro-integration