extracting-tables
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill documents the extraction of data from external, potentially untrusted sources such as PDFs, spreadsheets, and images. Maliciously crafted content within these files could attempt to influence the agent's behavior if the extracted output is processed without further validation.
- Ingestion points: The
xberg extractcommand and the associated Python/Node.js SDKs ingest local files (e.g.,report.pdf,workbook.xlsx,invoices). - Boundary markers: The instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings when the extracted data is passed to an LLM context.
- Capability inventory: The extracted content is typically rendered as Markdown or JSON, which is then used by the agent for downstream tasks like data analysis or reporting.
- Sanitization: No explicit sanitization or filtering steps for the extracted content are documented in the skill instructions.
Audit Metadata