The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by instructing the agent to process and translate content from external .strings files without proper sanitization or boundary markers. * Ingestion points: The agent reads translation content from Localizable.strings files as part of the Complete Localization Workflow and AI-Powered Translation steps. * Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the string values as untrusted data or to ignore any embedded commands. * Capability inventory: The skill includes scripts that can write to the local filesystem (sync_translations.py) and instructions to execute shell commands like grep and tuist. * Sanitization: The skill lacks any mechanism to verify that translated strings do not contain instructions designed to manipulate the agent's behavior during the translation process.

app-localization