The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill includes a command npx --yes submit-expo-feedback@latest in the feedback section. This command instructs the environment to download a package from the NPM registry at runtime. The package is not from a known trusted organization and is not pinned to a specific version, creating a supply chain risk.\n- [REMOTE_CODE_EXECUTION]: The execution of npx --yes submit-expo-feedback@latest involves downloading and running code from an external source. Because the source is unverified and uses the @latest tag, it could lead to arbitrary code execution if the package is compromised.\n- [COMMAND_EXECUTION]: The feedback command template npx ... "<actionable feedback>" allows for the interpolation of user-supplied content. This pattern creates a command injection vulnerability if the input contains shell metacharacters (e.g., backticks or semicolons) that are not properly sanitized before the command is executed by the agent or user.\n- [INDIRECT_PROMPT_INJECTION]: The skill documents a mechanism for passing serializable props from a native screen into a webview. This establishes a data flow that could be exploited if malicious content is processed by the webview's logic or passed back to native actions.\n
Ingestion points: Data enters the webview through component props defined in files like components/dom-component.tsx.\n
Boundary markers: None provided; the instructions do not suggest using delimiters or escape sequences for the props.\n
Capability inventory: The webview can execute JavaScript and invoke exposed native async functions (e.g., Alert.alert, saveData).\n
Sanitization: No input validation or sanitization routines are implemented in the provided code examples.

expo-dom