xdto-info
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill operates entirely on local files within a specified project directory. It reads XDTO package definitions (Package.bin) and summarizes their contents for the user. No network exfiltration, credential access, or malicious command execution was observed.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted metadata from XDTO files which could contain instructions intended to influence the agent.
- Ingestion points: Reads XML data from Package.bin and configuration files via the -PackagePath argument.
- Boundary markers: Output is structured with clear headers and bullet points representing the XDTO schema.
- Capability inventory: The skill performs file reads and writes (via the -OutFile parameter) using standard PowerShell and Python file operations. No network access or shell execution of ingested data is present.
- Sanitization: The skill utilizes standard XML parsing libraries (lxml in Python, .NET XmlDocument in PowerShell) which provide basic structural validation. The risk of the resulting summary causing the LLM to deviate from its instructions is low and typical for code analysis tools.
- [DYNAMIC_EXECUTION]: The script uses standard XML processing and string formatting to generate reports. It does not use eval(), exec(), or other dynamic execution sinks for user-provided or file-provided data.
Audit Metadata