The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill's internal utility scripts utilize dynamic module loading to resolve and execute internal components of its dependencies.
scripts/scan.mjs resolves the path to @ai-i18n/vite/internal/scan using require.resolve relative to a user-provided root directory and then imports it using pathToFileURL and import().
scripts/translations.mjs similarly resolves and dynamic-imports @ai-i18n/mcp/internal/agent based on a variable root path.
[COMMAND_EXECUTION]: The skill instructions provide several shell commands for installation, registration, and execution of its scanning and translation helpers.
Instructions in references/registration.md include commands to add skills, register MCP servers, and install plugins using npx and codex CLI tools.
The skill defines local scripts (scripts/scan.mjs, scripts/translations.mjs) that are intended to be executed by the agent to interact with the target application environment.
[EXTERNAL_DOWNLOADS]: The documentation guides the user and agent to install packages and plugins from external repositories and registries associated with the vendor.
references/registration.md contains instructions to download skills and the @ai-i18n/mcp@alpha package from GitHub and NPM.
The skill references documentation hosted at https://bosens-china.github.io/ai-i18n/llms.txt.
[INDIRECT_PROMPT_INJECTION]: The skill operates on untrusted data from the target application's source code and translation JSON files, creating a surface for indirect prompt injection.
Ingestion points: scripts/scan.mjs processes the application's entry points and dependencies; scripts/translations.mjs reads translation requests from JSON input or standard input.
Boundary markers: The documentation in references/plugin-hooks.md explicitly warns that diagnostics returned by plugins are untrusted data and should not be treated as instructions.
Capability inventory: The skill has capabilities to read filesystem content, execute internal scanning logic, and write to project translation and override files.
Sanitization: While the skill validates template tokens and schema structures for batch updates, it relies on the agent to handle natural language content which could contain embedded instructions.

use-ai-i18n-mcp