copilot-studio-agent-test
Pass
Audited by Gen Agent Trust Hub on Nov 7, 2026
Risk Level: SAFE
Full Analysis
- Execution of Bundled JavaScript: The skill performs its operations using a self-contained JavaScript bundle (
cs-agent-test.cjs). While bundled code can be less transparent than raw source, the author provides a comprehensive list of third-party notices and dependencies, and the script is provided in a readable format to facilitate security reviews. - Secure Configuration Management: Environment-specific settings and Entra ID values are stored in a dedicated directory within the user's home folder (
~/.cs-agent-test/). This is a recommended security practice that prevents sensitive configuration from being stored within the skill's own directory or version-controlled files. - Delegated Authentication: The skill utilizes a device-code authentication flow via the Microsoft 365 Agents SDK. This avoids the need for hardcoded client secrets or long-lived credentials, requiring interactive user sign-in and consent which aligns with modern identity security standards.
- Scoped Data Ingestion: The tool processes external Word (.docx) and CSV files provided by the user. While any file ingestion presents a potential surface for indirect prompt injection, the risk is minimized by the skill's specific focus on structured Q&A tables and the requirement for human-in-the-loop grading of the results.
- Use of Official SDKs: The skill's primary functionality is built upon the official
CopilotStudio.Copilots.Invokecapability. All network interactions are directed toward established Microsoft service endpoints, ensuring the skill operates within the expected ecosystem of the vendor.
Audit Metadata