sinch-in-app-calling

Pass

Audited by Gen Agent Trust Hub on Sep 29, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references and fetches the Sinch RTC SDK from official vendor distribution channels. This includes the JavaScript SDK from the Sinch CDN (cdn.sinch.com), the iOS SDK via the vendor's GitHub repository for Swift Package Manager, and the Android SDK via Maven Central or direct AAR downloads from download.sinch.com.
  • [COMMAND_EXECUTION]: Contains instructions for standard developer tools including Gradle (Android), npm/script tags (JavaScript), and Xcode/SPM (iOS) to facilitate SDK integration into user projects.
  • [PRIVILEGE_ESCALATION]: The integration guides for mobile platforms declare the necessity of standard VoIP and video calling permissions, such as INTERNET, RECORD_AUDIO, CAMERA, and READ_PHONE_STATE. These are documented as required for the primary functionality of the skill.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest user project context (such as platform, language, and build system) and external documentation from developers.sinch.com to guide the integration process.
  • Ingestion points: Platform detection from user project details and documentation retrieval from vendor-controlled domains (SKILL.md).
  • Boundary markers: Utilizes a clear instructional hierarchy and explicit "Source of Truth" rules to prioritize authoritative vendor documentation over navigational summaries.
  • Capability inventory: The skill provides code generation and documentation navigation; it does not execute arbitrary shell commands or code on user-supplied data.
  • Sanitization: The instructions guide the agent to verify implementation steps with the user before proceeding.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 29, 2026, 01:06 PM