unicode-tools
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input and external character metadata through its search and lookup tools, creating a surface for indirect instructions to enter the agent's context.
- Ingestion points: Input queries in
src/unicodefyi/engine.py(search function) and all tools insrc/unicodefyi/mcp_server.py(char_info,char_encode,unicode_search). - Boundary markers: Absent. The tools return Markdown-formatted strings without explicit delimiters or warnings to ignore embedded instructions.
- Capability inventory: The skill includes a REST API client in
src/unicodefyi/api.pywith network-read capabilities (viahttpx), but it does not have file-system write access or subprocess execution capabilities. - Sanitization: Absent. The skill returns character names and properties as retrieved from the Unicode standard (
unicodedata) or its internal mapping table without escaping or filtering.
Audit Metadata