meta-info
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill extracts and displays metadata from user-provided XML files, such as synonyms, comments, and object names. This creates an attack surface for indirect prompt injection where a malicious XML file could contain instructions hidden in these fields to influence the agent's subsequent actions.
- Ingestion points: XML metadata files read via the
-ObjectPathargument in bothscripts/meta-info.ps1andscripts/meta-info.py. - Boundary markers: The skill output lacks specific security delimiters or warnings to the agent to disregard instructions potentially embedded in the processed metadata.
- Capability inventory: The skill has read access to the local filesystem for XML parsing and outputs text to the agent's context; it does not perform network operations or execute arbitrary system commands derived from the XML content.
- Sanitization: The scripts extract raw text from XML nodes (e.g., via the
get_ml_textfunction) and display it without filtering or escaping content for potential prompt injection patterns.
Audit Metadata