The Agent Skills Directory

[COMMAND_EXECUTION]: Several CLI commands described in the documentation, such as build, watch, editor, and live, support a --with argument. This flag allows the execution of arbitrary shell commands in parallel with the Intlayer process (e.g., npx intlayer build --with "next dev").
[CREDENTIALS_UNSAFE]: The login command initiates a local HTTP server on a random port to capture sensitive credentials (Client ID and Client Secret) from the Intlayer CMS via the browser. Additionally, the documentation instructs users on how to manage these secrets using environment variables and provides an --api-key flag for direct integration with AI providers.
[INDIRECT_PROMPT_INJECTION]: The fill, doc-translate, and doc-review commands ingest local project content (dictionaries and documentation files) to process them through external AI services (OpenAI, DeepSeek). This creates an attack surface where malicious instructions embedded in the processed files could potentially influence the AI's output or the agent's behavior.
Ingestion points: Reads src/**/*.content.{ts|js|mjs|cjs|json} and documentation files (.md, .mdx).
Boundary markers: No explicit boundary markers or sanitization steps are documented for the content sent to AI providers.
Capability inventory: File system read/write, network communication (API calls), and parallel command execution.
Sanitization: Static documentation does not mention content escaping or instruction filtering before processing.
[PRIVILEGE_ESCALATION]: The init command programmatically modifies project-level configuration files, including .gitignore, tsconfig.json, vite.config.ts, and next.config.js. While intended for setup, this capability allows the tool to alter the project's build and security settings.

intlayer-cli