invoice-processor
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill transmits invoice data to the Zhipu AI (bigmodel.cn) API endpoint for processing. This is the primary intended function of the skill and is explicitly documented for the user in the documentation and setup instructions. The skill uses standard, well-known Python libraries (aiohttp, PyMuPDF, openpyxl) for its operations.
- [INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests untrusted external data in the form of invoice images and PDFs which could contain text instructions designed to steer the AI's extraction behavior.
- Ingestion points: The
scripts/invoice_ocr.pyscript reads files from a user-specified directory. - Boundary markers: There are no explicit text-based boundary markers for the image content processed by the vision model.
- Capability inventory: The skill can read local files, perform network requests to the Zhipu AI API, write Excel files, and delete temporary processing files.
- Sanitization: The script attempts to parse specific JSON structures from the model's output to validate the extracted data.
Audit Metadata