localization-toolkit
Pass
Audited by Gen Agent Trust Hub on Nov 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (application source code and JSON locale files) to audit and replace strings. It lacks explicit boundary markers or sanitization for instructions that might be embedded within the project code or translation files.
- Ingestion points: The
scripts/i18n_audit.pyscript (via--srcand--localearguments) andrgsearch commands process files within the target project directory. - Boundary markers: Absent; the skill treats all file content as valid data for processing without delimiters to prevent the agent from obeying embedded instructions.
- Capability inventory: The skill performs file system reads, executes shell commands (
rg), runs local Python scripts, and initiates package installations. - Sanitization: No escaping or validation is performed on strings extracted from or written to the codebase.
- [INDIRECT_PROMPT_INJECTION]: The skill contains a file named
.security-scan-passedthat claims a security scan was successful and provides a content hash. This is a self-referential pattern designed to influence the verdict of security analysis tools or reviewers, which is a known adversarial technique.
Audit Metadata