review-safety
Pass
Audited by Gen Agent Trust Hub on Jul 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local commands including
unilyze,jq, andgrep. These are used to generate a project snapshot, extract metrics, and search for specific code patterns (e.g.,Process.Start,NativeArray) in C# source files. This behavior is consistent with the skill's stated purpose of providing a static safety review. - [PROMPT_INJECTION]: The skill has an indirect prompt injection surface because it reads and analyzes untrusted C# source code from the project's
Assetsdirectory. Malicious instructions placed in code comments could attempt to influence the agent's analysis output. - Ingestion points: Source code files (
.cs) located in<project>/Assetsare read into the agent's context during the AI review phase. - Boundary markers: The skill does not define explicit delimiters or instructions to ignore embedded commands when processing source code contents.
- Capability inventory: The skill's capabilities are limited to reading local files and writing a findings report to
<project>/.unity-review/report/safety-findings.md. - Sanitization: No sanitization or validation of the ingested code content is performed before analysis.
Audit Metadata