The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes strings within application source code, which serves as an ingestion point for untrusted data. * Ingestion points: The skill reads various source file formats including .tsx, .jsx, .vue, .svelte, .html, and .ts. * Boundary markers: There are no specific boundary markers or instructions to the agent to disregard instructions that might be contained within these application strings. * Capability inventory: The skill uses shell commands like grep to read files and is intended to perform file writes for extraction and translation. * Sanitization: Instructions are provided to preserve placeholder syntax and brand names, but there is no mechanism for sanitizing the semantic content of the extracted strings against potential prompt injection.
[COMMAND_EXECUTION]: The skill uses grep for local codebase auditing. * Evidence: Shell commands in SKILL.md and references/audit-patterns.md use grep to locate hardcoded strings, toast notifications, and validation errors. These commands are restricted to the local source directory.
[SAFE]: The skill uses industry-standard libraries and practices. * Evidence: Recommendations for i18next, react-i18next, vue-i18n, and next-intl involve well-known packages from the official npm registry. The skill also explicitly references security advisories (CVE-2026-48714 and CVE-2026-48713) to ensure users pin safe versions of middleware.

localize