datasheets
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from component datasheets, creating a risk that malicious text within a PDF could influence agent behavior.
- Ingestion points: PDF files are read and processed using subagent templates in the prompts/ directory (e.g., prompts/base.md, prompts/pinout.md).
- Boundary markers: Subagent instructions lack explicit delimiters or instructions to disregard potential commands or instructions found within the PDF content.
- Capability inventory: Includes local command execution via subprocess (scripts/datasheet_page_selector.py) and filesystem write access for caching extracted data (scripts/merge_results.py).
- Sanitization: No sanitization or filtering is applied to the text content extracted from the PDFs before processing.
- [COMMAND_EXECUTION]: The skill uses local system utilities to process PDF files.
- Evidence: scripts/datasheet_page_selector.py uses subprocess.run to execute pdfinfo and pdftotext. These calls use fixed argument lists and are intended for metadata and text extraction purposes.
Audit Metadata