The Agent Skills Directory

[SAFE]: No malicious patterns or attack vectors were detected. The skill is focused on standard developer workflows for internationalization.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves processing application source code to extract translatable strings and generating translations via LLM. This creates an attack surface where malicious content in source code could theoretically influence the agent during extraction or translation steps.
Ingestion points: Application source files including .tsx, .jsx, .vue, .svelte, .ts, and .html are read using grep patterns to identify user-facing strings.
Boundary markers: The skill provides a robust prompt template in references/translation-quality.md that includes explicit preservation rules and constraints to prevent the LLM from deviating from its translation task.
Capability inventory: The skill uses grep for read operations and requires file system writes to update locale catalogs and application source files.
Sanitization: The documentation specifically warns against using manual .replace() calls on translated output, noting it as a risk that bypasses escaping mechanisms, and enforces the use of established i18n library interpolation.
[EXTERNAL_DOWNLOADS]: The skill recommends several well-known Node.js packages (e.g., i18next, react-i18next). These references include specific version pins and security floors, citing GitHub Security Advisories (GHSAs) to ensure the developer uses versions patched against known prototype pollution vulnerabilities.

i18n-localization