godot-headless-workflow
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute several shell commands, including
godot --headlessfor resource importing, syntax checking (--check-only), and running the application for smoke testing. It also usesgit difffor change detection andkeytoolfor managing Android keystores.- [INDIRECT_PROMPT_INJECTION]: The workflow involves the execution of project-local logic via the Godot engine, specifically during smoke tests (godot --headless --quit-after 300) and unit testing withgdUnit4. This creates an execution surface where a malicious project could run arbitrary code when the agent follows these development instructions. - Ingestion points: The workflow processes
.gdscript files and.tscnscene files from the repository. - Boundary markers: No explicit delimiters or instructions are provided to sanitize or inspect project-local code before execution.
- Capability inventory: The agent uses the
bashtool to run the Godot engine, which has the capability to execute the GDScript logic contained within the project. - Sanitization: There is no mention of code sanitization or sandboxing for the execution of tests.- [EXTERNAL_DOWNLOADS]: The documentation references external third-party plugins, specifically
gdUnit4and theGodotx RevenueCatplugin, hosted on GitHub. These references are informational and do not involve automated execution of untrusted remote content.- [DYNAMIC_EXECUTION]: The skill provides templates for GDScript utility scripts (e.g.,smoke_test.gdandbuild_scene.gd) that are intended to be written to the project and executed via the Godot command line. These scripts perform operations like instantiating scenes or programmatically generating resource files from templates provided in the instructions.
Audit Metadata