test-setup
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEDYNAMIC_CONTEXT_INJECTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [DYNAMIC_CONTEXT_INJECTION]: The skill utilizes the
!command`` syntax in the frontmatter to execute a shell script (yaml-helper.sh) at skill load time. This command resolves the automation mode by reading project configuration files. - [INDIRECT_PROMPT_INJECTION]: The skill's behavior is influenced by the contents of
project.yamland.claude/docs/technical-preferences.md. While this is intended for engine detection, it represents an ingestion point for untrusted data. - Ingestion points:
project.yaml,.claude/docs/technical-preferences.md. - Boundary markers: Absent; the skill procedurally reads specific keys from these files.
- Capability inventory: File writing (
Write), file reading (Read), and shell execution (Bash) are used to create the scaffold. - Sanitization: The skill does not explicitly sanitize the engine name or version strings read from the configuration files before using them in generated file content.
- [EXTERNAL_DOWNLOADS]: The generated CI/CD workflows (.github/workflows/tests.yml) reference third-party GitHub Actions for Godot (
MikeSchulze/gdUnit4-action) and Unity (game-ci/unity-test-runner). These are standard for their respective communities but represent external dependencies in the resulting project code. - [COMMAND_EXECUTION]: The skill uses the
Bashtool and dynamic context injection to execute shell commands for configuration management. The scope of these commands is restricted to the project's helper scripts as defined in the frontmatter.
Audit Metadata