The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to use npx testflight as a shortcut for building and submitting to TestFlight. This command attempts to download and execute a package named 'testflight' from the npm registry. As this is not an official Expo tool or a verified utility from a known developer, it presents a risk of executing arbitrary code if an attacker publishes a malicious package under that name.
[DATA_EXFILTRATION]: The skill describes workflows involving highly sensitive credentials, specifically requiring the agent to manage paths for Google Service Account keys (google-service-account.json) and Apple Distribution API keys (.p8 files). While the skill suggests using environment variables for CI/CD, the instruction for the agent to directly interact with these credential files on the local file system creates an exposure risk.
[PROMPT_INJECTION]: The skill creates an attack surface for indirect prompt injection through its metadata management functionality. It allows the agent to ingest application metadata from external sources which is then pushed to app stores.
Ingestion points: Metadata is read from store.config.json and external localization APIs via fetch in dynamic JavaScript configurations.
Boundary markers: None are present; the instructions do not include delimiters or warnings to ignore embedded instructions within the metadata.
Capability inventory: The skill possesses capabilities for subprocess execution (eas-cli), network access (fetch), and file manipulation.
Sanitization: There is no mention of validating or sanitizing the external metadata before it is processed and uploaded.
[EXTERNAL_DOWNLOADS]: In the app-store-metadata.md reference, the skill provides a code example that fetches application copy from an external URL (https://api.example.com/app-store-copy). This promotes a pattern where production deployment configurations are dynamically retrieved from unverified remote endpoints.

expo-deployment