The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill describes how to create native bridge functions that accept untrusted input from the JavaScript environment, which can lead to indirect prompt injection vulnerabilities if the AI agent processes data via these modules.
Ingestion points: Arguments for Function and AsyncFunction blocks, as well as Prop setters in native views (e.g., references/native-module.md and references/native-view.md).
Capability inventory: The documentation includes examples of high-privilege operations such as reading arbitrary URLs or local files using Data(contentsOf: url) in Swift and url.readText() in Kotlin.
Boundary markers: The provided code templates do not include delimiters or safety instructions to distinguish between data and instructions when processing external inputs.
Sanitization: No input validation, escaping, or URI sanitization is demonstrated in the boilerplate code snippets.
[COMMAND_EXECUTION]: The skill instructs the user to execute shell commands to scaffold and manage native modules.
Evidence: Recommended use of npx create-expo-module@latest for module initialization and npx expo prebuild for project synchronization in references/create-expo-module.md and references/config-plugin.md.
[DATA_EXPOSURE]: The guides demonstrate patterns for reading sensitive configuration values directly from native application storage.
Evidence: references/config-plugin.md provides code snippets for extracting values from the iOS Bundle and Android PackageManager metadata (e.g., MY_API_KEY).

expo-module