The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides specific bash commands using grep to search the local filesystem for GDScript patterns, including potential physics-related process calls and hardcoded node paths in the scripts/ directory.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze Godot scripts while possessing shell and file access capabilities, creating a surface for indirect instructions to influence agent behavior.
Ingestion points: External GDScript (.gd) and C# (.cs) files are read into the agent's context via bridge tools like script_read and shell commands.
Boundary markers: There are no explicit delimiters or instructions telling the agent to treat the content of analyzed scripts as untrusted data or to ignore instructions contained within comments.
Capability inventory: The skill requests bash and read permissions and includes tools for file search, directory browsing, and script validation.
Sanitization: No sanitization or filtering logic is present to prevent malicious instructions inside analyzed scripts from being interpreted by the agent.

godot-code-review