zh-en-translation-polish
Pass
Audited by Gen Agent Trust Hub on Aug 9, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is designed to ingest and process arbitrary Chinese text for translation, which constitutes a surface for indirect prompt injection attacks.
- Ingestion points: The agent receives external text either via direct prompt input or by reading user-specified files in
SKILL.md. - Boundary markers: The skill employs Markdown blockquote delimiters (
>) to separate Chinese source text from English translations. - Capability inventory: The skill utilizes
Read,Write, andBashtools to manipulate files and execute a local diagnostic script (chinglish_scan.py). - Sanitization: No explicit sanitization or filtering of the source text for malicious instructions is described beyond the translation logic itself.
- [COMMAND_EXECUTION]: The skill generates and executes a short Python script at runtime to derive English-only documents from bilingual sources.
- Evidence:
SKILL.mdcontains a Python one-liner that filters lines starting with>and writes the output to a new file. - Context: This is a localized, static script generation used solely for document formatting and does not incorporate untrusted external input into the execution logic.
Audit Metadata