powershell-safe-invocation
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides instructions for executing native commands and PowerShell cmdlets using secure patterns like array-based arguments and hashtable splatting. These methods are designed to prevent shell injection vulnerabilities.
- [DYNAMIC_EXECUTION]: The skill instructs the agent to create and run temporary .ps1 files for complex multiline operations, which avoids the risks associated with multi-layered shell escaping and quoting.
- [INDIRECT_PROMPT_INJECTION]: The skill identifies potential injection vectors in path processing and provides a mandatory evidence chain: 1) Ingestion points: Path processing in SKILL.md (e.g., $candidate) and reference.md (e.g., Get-Content -LiteralPath $path). 2) Boundary markers: The skill recommends using array-based argument passing (@argList) and splatting (@params) to maintain clear boundaries. 3) Capability inventory: Native executable invocation via the & operator and script execution via pwsh.exe -File are detailed in both files. 4) Sanitization: Explicit validation steps are provided, such as checking path prefixes and using -LiteralPath to prevent wildcard interpretation.
- [SAFE]: The skill is entirely educational and focuses on improving the security posture of the agent's shell interactions.
Audit Metadata