The Agent Skills Directory

[PROMPT_INJECTION]: None. The skill uses instructional language for code generation without attempting to bypass safety filters.
[DATA_EXPOSURE_AND_EXFILTRATION]: None. No hardcoded credentials or sensitive file access patterns were found. The skill advises keeping sensitive keys on the native side and redacting sensitive headers in monitoring tools.
[OBFUSCATION]: None. No hidden characters, homoglyphs, or encoded commands were detected.
[REMOTE_CODE_EXECUTION]: None. The skill discusses Over-The-Air (OTA) updates using signed releases and GitHub, which are standard deployment mechanisms for this library's use case. It emphasizes verification of signed assets.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external inputs through various "bridges" (e.g., Hubs, NFC, Barcodes). It provides explicit guidance for developers:
Ingestion points: External data enters through HubCall, BridgeQuery, BridgeBody, and NDEF records.
Boundary markers: The system uses AppDeviceBridgePolicies and the WebView session token to isolate bridge access.
Capability inventory: The library has capabilities for file system access (IFilesBridge), network operations (HttpTransfers), and system control (VPN, Biometrics).
Sanitization: The documentation warns that hub arguments must be treated as untrusted input from other devices.
[PRIVILEGE_ESCALATION]: None. Permission requests mentioned (camera, location, biometrics, Bluetooth) are standard for mobile/desktop app functionality and require OS-level user consent and manifest declarations.

shiny-appdevicebridge