azure-avatar-integrate
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill is installed using a vendor-provided command (
npx skills add) that fetches components from the official Microsoft Foundry repository on GitHub (github.com/microsoft-foundry/voicelive-samples).- [COMMAND_EXECUTION]: The helper scriptscripts/feedback_form_url.pyis used to launch the user's default web browser viawebbrowser.openoros.startfileto facilitate the optional feedback process.- [INDIRECT_PROMPT_INJECTION]: The skill identifies and manages a potential indirect prompt injection surface associated with summarizing integration issues and generating feedback reports. - Ingestion points: Integration status, symptoms, and resolution history are stored in a local JSON record (
.azure-avatar-integrate/<integration-id>.json) which is later processed to generate a feedback draft. - Boundary markers: The process requires a mandatory draft review by the developer and an explicit consent confirmation before any external URL is generated or opened.
- Capability inventory: The skill performs local file system writes for issue tracking, executes a local Python helper script, and invokes the system's default browser.
- Sanitization: The
feedback_form_url.pyscript incorporates a comprehensive security filter using regular expressions to detect and block the inclusion of credentials (API keys, bearer tokens), internal IP addresses, and private local file paths in the generated feedback URL.- [SAFE]: The skill instructions proactively mitigate data exposure risks by explicitly prohibiting the AI agent from requesting or reading secret values, instructing that all service keys and identity tokens must remain secured within the application's backend environment.
Audit Metadata