pdf-table-data-conversion
Pass
Audited by Gen Agent Trust Hub on Aug 30, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [Indirect Prompt Injection Surface]: The skill processes external PDF documents, which are untrusted data sources. Maliciously crafted content within a PDF could potentially attempt to influence the agent's instructions, particularly during manual review steps.
- Ingestion points: External PDF files retrieved from knowledge sources or user uploads (referenced in SKILL.md and processed by scripts/extract_tables.py).
- Boundary markers: The instructions do not define specific delimiters to isolate processed PDF data from agent instructions.
- Capability inventory: The skill executes a Python script for file reading and writing, and the agent has the ability to read document text directly.
- Sanitization: The extraction script proactively prevents Excel Formula Injection by explicitly setting cell data types to strings.
- [Local Script Execution]: The skill utilizes a bundled Python script (scripts/extract_tables.py) to perform deterministic data extraction. This is a standard pattern for specialized data processing within the agent's environment.
Audit Metadata