The Agent Skills Directory

[COMMAND_EXECUTION]: The skill creates and executes shell scripts (i18n-push.sh, i18n-pull.sh) to interface with the Lokalise CLI for translation management. It also utilizes child_process.exec and execSync within TypeScript file watchers to automate synchronization tasks. These actions are standard for development automation and localized to the project environment.
[DYNAMIC_EXECUTION]: Employs dynamic import() statements to lazy-load localization files in React and Vue application code. It also generates and runs local scripts for monitoring file system changes. These dynamic operations are confined to defined project paths.
[INDIRECT_PROMPT_INJECTION]: The skill handles translation data from external sources (Lokalise) stored in local JSON files. This creates a surface where malicious or unexpected content in translation strings could potentially influence agent behavior if processed as instructions.
Ingestion points: Translation files located at src/locales/*.json which are read by both the agent and synchronization scripts.
Boundary markers: No explicit delimitation or 'ignore instructions' markers are used when processing the locale data.
Capability inventory: The skill possesses file system access and shell command execution capabilities.
Sanitization: There is no dedicated sanitization or validation of the content within the translation files.
[EXTERNAL_DOWNLOADS]: Utilizes npx to invoke tools like tsx and husky, which involves retrieving packages from the npm registry. These are well-established utilities within the Node.js development ecosystem.

lokalise-local-dev-loop