The Agent Skills Directory

[COMMAND_EXECUTION]: The skill exposes powerful tools such as execute_editor_script and execute_game_script which allow for the execution of arbitrary GDScript. This provides a direct path to the underlying host system via Godot's built-in APIs (e.g., the OS class), enabling actions such as file system modification, process execution, and network operations.
[REMOTE_CODE_EXECUTION]: The installation process requires cloning code from an external GitHub repository to install the Godot plugin, and the server component is distributed as a third-party package. These external sources are responsible for the core logic that the AI interacts with.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it is designed to read and process untrusted content from the local filesystem, editor logs, and error reports. This data can influence the AI's behavior and potentially trigger the execution of malicious GDScript via the toolset.
Ingestion points: read_script, get_output_log, get_editor_errors, search_in_files.
Boundary markers: No explicit boundary markers or "ignore embedded instructions" warnings are documented for the data being ingested.
Capability inventory: execute_editor_script, execute_game_script, edit_script, create_script, attach_script, simulate_key.
Sanitization: No sanitization or validation of the ingested external content is mentioned before it is processed by the AI.

godot-mcp-pro-integration