The Agent Skills Directory

[COMMAND_EXECUTION]: The 'content.formatCommand' configuration option allows users to define shell commands (e.g., 'npx prettier --write "{{file}}"') that are executed on local dictionary files. This is a legitimate feature for build tools but represents an execution surface.
[EXTERNAL_DOWNLOADS]: The configuration examples utilize 'npx' (e.g., for 'prettier', 'biome', or 'eslint'), which may download and execute packages from the npm registry if they are not already installed locally.
[PROMPT_INJECTION]: The skill facilitates AI-powered translation and auto-filling via the 'ai' configuration. This introduces a surface for indirect prompt injection if source dictionary files contain instructions that the AI model might obey. * Ingestion points: Source content files (.content.ts, .content.js, .content.json) and the 'applicationContext' configuration field. * Boundary markers: No specific delimiters or 'ignore' instructions are mentioned in the schema to isolate untrusted data from AI instructions. * Capability inventory: AI-powered translation filling, local file writes via formatting commands, and synchronization with the Intlayer CMS. * Sanitization: The configuration does not explicitly define sanitization or escaping mechanisms for dictionary content before AI processing.

intlayer-config