godot-pipeline-automation
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches the gdUnit4 testing framework from its official GitHub repository. The installation process includes mandatory SHA-256 checksum verification against a hardcoded known-good hash to protect against supply chain attacks.
- [EXTERNAL_DOWNLOADS]: Configures CI/CD workflows to download official Godot Engine binaries and export templates from the primary Godot project releases. The generated scripts verify these downloads using SHA-512 sums.
- [COMMAND_EXECUTION]: Utilizes subprocesses to perform essential development tasks, such as linting GitHub Action workflows via actionlint and initiating a project-local Model Context Protocol (MCP) server for agent-based scene interaction.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted asset files (GLB) and manifests. It contains specific security logic in
scripts/agentkit/pipeline/glb_probe.gdandscripts/gd_pipeline.pyto detect and block assets containing external URIs, which prevents the engine from making unauthorized network requests or leaking metadata during the import process. - [PRIVILEGE_ESCALATION]: Applies executable permissions (
chmod +x) to CI scripts and downloaded binaries. This is a standard and necessary operation for build engineering automation. - [DYNAMIC_EXECUTION]: Includes a C-based GDExtension example for smoke-testing the native code loading chain. This is used for verification of the build environment rather than runtime execution of untrusted code.
Audit Metadata