mena-product-localization
Pass
Audited by Gen Agent Trust Hub on Jun 30, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references documentation and libraries from trusted organizations such as Google (libphonenumber, libaddressinput), Unicode (CLDR, Bidi), W3C (i18n, ALReq), and Amazon (Payment Services). It also includes links to official government services in Saudi Arabia (ZATCA, SDAIA), the UAE (UAE Pass, MoEC), and Egypt for regional compliance context.
- [COMMAND_EXECUTION]: Includes utility scripts (
rtl_copy_lint.js,locale_format_examples.js) used for deterministic linting of RTL content and generating locale-specific examples. These scripts operate locally and do not involve remote code execution or network operations. - [DATA_EXFILTRATION]: Strong data safety rules are implemented across
SKILL.mdand reference files. The skill explicitly forbids the processing of live payment data and instructs agents to protect PII like phone numbers, national IDs, and physical addresses by using masking and avoiding real data in examples. - [INDIRECT_PROMPT_INJECTION]: The skill recognizes the risk of ingesting untrusted user content during localization reviews. It mitigates this by instructing the use of Unicode bidirectional isolation (using
bdianddirattributes) inreferences/bidi-text.mdto prevent content from disrupting UI layout or leaking instructions.
Audit Metadata