applicationinsights-web-ts
Pass
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill sets up instrumentation to ingest data from the web environment, including page views, clicks, and AI agent messages.
- Ingestion points: Browser telemetry, user interactions, and GenAI model inputs/outputs processed within the application.
- Boundary markers: The documentation suggests using telemetry initializers to filter and drop specific items based on logic.
- Capability inventory: The SDK enables tracking of events, exceptions, and dependencies, but does not grant the agent shell execution or file system write access on the host.
- Sanitization: Explicitly provides patterns for scrubbing secrets from URLs (e.g., tokens, keys) and warns against capturing PII or sensitive GenAI content without authorization.
- [EXTERNAL_DOWNLOADS]: The skill recommends downloading the core SDK from Microsoft's official CDN and installing NPM packages from the Microsoft organization.
- Evidence: References to https://js.monitor.azure.com/scripts/b/ai.3.gbl.min.js and @microsoft/applicationinsights-web family of packages.
- [CREDENTIALS_UNSAFE]: The skill provides instructions for managing connection strings using environment variables (e.g., VITE_APPINSIGHTS_CONNECTION_STRING), which is a standard and recommended practice for public-facing web clients.
Audit Metadata