expo-native-ui
Fail
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: HIGHCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The
SKILL.mdfile contains instructions for the agent to execute the shell commandnpx --yes submit-expo-feedback@latestto submit feedback. This involves running code from a remote source directly in the user's environment. - [REMOTE_CODE_EXECUTION]: The use of
npx --yes ...@latestautomatically fetches and executes the most recent version of a package from the npm registry. This behavior bypasses version pinning and can be used to execute arbitrary code if the package name is typosquatted or the original package is hijacked. - [INDIRECT_PROMPT_INJECTION]: The skill establishes a vulnerability surface where untrusted data could influence agent behavior or exploit its capabilities.
- Ingestion points: Media files (photos, audio, video) processed via
expo-cameraandexpo-image-pickerinreferences/media.md, and data retrieved fromlocalStorageorexpo-sqliteinreferences/storage.md. - Boundary markers: None detected; there are no instructions to the LLM to treat data from these sources as untrusted or to use specific delimiters.
- Capability inventory: Shell command execution via
npx, persistent file system writes viaexpo-file-system, and hardware access including camera and microphone permissions. - Sanitization: No evidence of input validation, escaping, or sanitization for content retrieved from external files or storage before it is processed or used in logic.
Recommendations
- AI detected serious security threats
Audit Metadata