notes-cram
Pass
Audited by Gen Agent Trust Hub on Mar 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides a specific Python script within
SKILL.mdto be executed for extracting text from PDF files using thefitz(PyMuPDF) library. Additionally, it specifies the execution of the shell commandpip install pymupdf -qto ensure the required environment is set up.- [EXTERNAL_DOWNLOADS]: The skill initiates an external download of thepymupdfpackage from the Python Package Index (PyPI) to support its PDF parsing functionality.- [PROMPT_INJECTION]: The skill exhibits a high susceptibility to indirect prompt injection as its primary role is to ingest and summarize untrusted external data such as user-provided PDFs, audio transcripts, and video subtitles. - Ingestion points: Data enters the agent context via the PDF extraction script in
SKILL.mdand the multi-media processing instructions inprompt_補習班名師.md. - Boundary markers: Absent. The prompts do not define specific delimiters or instructions to prevent the agent from obeying commands embedded within the input materials.
- Capability inventory: The skill possesses the ability to execute Python code and perform file system read operations.
- Sanitization: None. There is no evidence of filtering or validation of the content extracted from external sources before it is processed by the model.
Audit Metadata