The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documentation describes a formatCommand configuration option in references/concept_configuration.md. This allows the execution of arbitrary shell commands (e.g., npx prettier --write "{{file}}") whenever content files are created or updated. If an attacker can influence this configuration, they could achieve command injection.
[DYNAMIC_EXECUTION]: The Intlayer compiler, described in references/concept_configuration.md, has a saveComponents setting. When enabled, this allows the tool to permanently rewrite application source code files on disk after transformation. This represents a powerful file system modification capability.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of external content files (.content.ts, .content.json, etc.) for AI-powered translation, creating a surface for indirect prompt injection attacks.
Ingestion points: The skill scans directories (defined in content.contentDir) for dictionary files that contain translatable text.
Boundary markers: No explicit instructions are provided to the agent to treat translatable content as untrusted or to ignore embedded instructions within those files.
Capability inventory: The skill has the ability to execute shell commands via formatCommand and write/rewrite files via the compiler and saveComponents features (found in references/concept_configuration.md).
Sanitization: There is no mention of sanitizing or escaping the content before it is processed by the AI for translation or generation.

intlayer-config