The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, specifically translation files loaded via dynamic imports and the Lokalise API. This content is interpolated into the application's UI, which creates a potential surface for indirect prompt injection if the source data is compromised. Ingestion points: messages/*.json files and the sync-translations.ts script. Boundary markers: Absent. Capability inventory: Filesystem writes (fs.writeFile), network requests (fetch), and package installation (npm install). Sanitization: Absent.\n- [EXTERNAL_DOWNLOADS]: The skill provides a script to download translations from the well-known and trusted service Lokalise (api.lokalise.com).\n- [COMMAND_EXECUTION]: Instructions include standard package management commands like npm install next-intl to set up the localization environment.\n- [DYNAMIC_EXECUTION]: The implementation utilizes dynamic imports for loading locale-specific message files. This is the standard architectural pattern for the next-intl library in Next.js applications.

localization-engineer