xberg-batch-extraction
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of content from over 107 different document formats, which are often used as input for LLMs. This creates a broad attack surface for indirect prompt injection attacks where malicious instructions are embedded within the processed data.
- Ingestion points: The
xberg batchandxberg extractcommands (documented inSKILL.mdandreferences/cli-reference.md) read local files and remote URLs. - Boundary markers: The configuration supports
insert_page_markers(documented inreferences/configuration.md), but these are intended for document structure rather than security delimitation. - Capability inventory: The library and CLI tools have extensive capabilities including file system access, network operations for URL ingestion, and OCR processing documented across all reference files.
- Sanitization: There are no explicit instructions or built-in mechanisms described for sanitizing or escaping extracted content to prevent instruction hijacking.
Audit Metadata