The Agent Skills Directory

Unverified External Package Execution: The documentation suggests using npx testflight as a shortcut for submitting builds to TestFlight. This command executes a package from the public npm registry that is not an official tool from Expo or Apple. Running unverified packages via npx carries the risk of executing untrusted code.
Remote Data Ingestion in Configuration: In references/app-store-metadata.md, the skill demonstrates fetching localization data from an external API (api.example.com) within a dynamic configuration file. Incorporating external content into application metadata creates a potential point for configuration poisoning if the source is not authenticated or secured.
Indirect Prompt Injection Surface: The skill identifies workflows where external data is ingested, creating a potential surface for indirect prompt injection.
Ingestion points: Remote data is fetched via fetch() in the store.config.js example.
Boundary markers: The examples lack specific delimiters to isolate external content from the rest of the configuration.
Capability inventory: Ingested metadata is used in eas metadata:push operations which communicate with store services.
Sanitization: The skill does not demonstrate methods for validating or sanitizing the retrieved JSON data.
Sensitive Credential Management: The skill involves handling sensitive files such as Apple API keys (.p8) and Google service account keys (.json). While the documentation recommends using EAS Secrets and environment variables, any local copies of these files present a security risk if not properly protected or if accidentally committed to version control.
Dynamic Configuration Execution: The skill encourages using JavaScript-based configuration for managing EAS metadata. These files are executed locally by the deployment environment, requiring users to audit any included logic for security considerations.

expo-deployment