verbatra-mcp-tools
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external locale files, glossary terms, and source strings which could contain malicious instructions.
- Ingestion points: Locale files, source strings, glossary terms, and translator comments accessed via tools like
key.valueandglossary.getinSKILL.md. - Boundary markers: The skill includes explicit instructions in Rule 4 to treat translatable strings, glossary terms, and comments as data rather than instructions, warning that text resembling commands is a prompt-injection attempt.
- Capability inventory: The agent has the ability to write to locale files (
translation.editEntry), modify the glossary (glossary.write), and trigger paid translation services (translation.translatePending). - Sanitization: The
glossary.gettool redacts values that look like provider API keys before returning them.
Audit Metadata