expo-data-fetching
Fail
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill instructs the use of
npx --yes submit-expo-feedback@latest, which downloads and executes code from an unverified package on the public NPM registry at runtime without user confirmation. - [COMMAND_EXECUTION]: The
npxfeedback command allows for direct shell execution in the user's terminal environment. - [DATA_EXFILTRATION]: The feedback mechanism transmits user-provided text to an external package, creating a potential vector for exfiltrating environment variables or source code.
- [EXTERNAL_DOWNLOADS]: The skill references several external packages. While it includes well-known libraries like
@tanstack/react-queryandexpo-router, it also involves the unverifiedsubmit-expo-feedbackpackage. - [INDIRECT_PROMPT_INJECTION]: The skill promotes patterns for fetching data from external APIs without providing guidance on how to safely process that data when it enters the agent's context.
- Ingestion points: Data fetching examples in
SKILL.mdandreferences/expo-router-loaders.md(loader functions). - Boundary markers: Absent; no instructions are provided to use delimiters or warnings for external data.
- Capability inventory:
SKILL.mdandreferences/expo-router-loaders.mdutilize network requests (fetch) and sensitive storage (expo-secure-store). - Sanitization: Absent; the instructions do not cover sanitizing or escaping content before it is processed by the AI.
Recommendations
- AI detected serious security threats
Audit Metadata