japanese-tutor

Pass

Audited by Gen Agent Trust Hub on Nov 10, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data via PDF and DOCX parsing scripts.
  • Ingestion points: Untrusted data enters the agent context through scripts/parse_pdf_gemini.py and scripts/parse_docx.py when processing user-provided study materials or homework.
  • Boundary markers: Absent. The SKILL.md instructions do not define clear delimiters or instruct the agent to disregard instructions found within the parsed content.
  • Capability inventory: The skill has the ability to read files, append content to local markdown reference files (references/vocab.md, references/grammar.md), and make network requests to the Gemini API.
  • Sanitization: Absent. There is no evidence of filtering or escaping logic for the text extracted from documents before it is interpolated into the agent's context or stored in the reference files.
  • [COMMAND_EXECUTION]: The skill utilizes several Python scripts for its core logic, including scripts/greet.py for dynamic greetings and the parsing scripts mentioned above. These scripts are used for their intended purposes and do not execute arbitrary shell commands or untrusted code.
  • [EXTERNAL_DOWNLOADS]: The scripts/parse_pdf_gemini.py script communicates with Google's Gemini API (google-generativeai) to perform multimodal analysis of PDF documents. This involves uploading document data to a well-known and established cloud service for processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 10, 2026, 10:07 PM