The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides code templates where untrusted user input is interpolated into prompts sent to external AI services and parameters for weather APIs.
Ingestion points: In app/api/ai+api.ts, the prompt variable is extracted directly from the request JSON. In app/api/weather+api.ts, the city is extracted from the query parameters.
Boundary markers: The provided code fragments do not use delimiters or boundary markers to isolate untrusted data from instructions.
Capability inventory: The skill demonstrates outbound network requests to OpenAI and weather services using this unvalidated data.
Sanitization: While the prose guidelines advise sanitization, the code snippets lack implementation, which could allow malicious data to influence the behavior of the upstream API.
[EXTERNAL_DOWNLOADS]: The skill references the installation and use of official tools and libraries from the Expo and Turso ecosystems.
Evidence: Recommends installing eas-cli globally via npm and utilizing the @libsql/client/web package for edge database access. These are well-known resources from recognized technology providers.

expo-api-routes