make-ai-teammate
Pass
Audited by Gen Agent Trust Hub on Oct 19, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- External Downloads: The skill clones starter samples from the official Microsoft Agent365-Samples repository on GitHub. This is a standard scaffolding step and uses a well-known, trusted source to provide developers with working templates.
- Command Execution: The skill executes common development and deployment commands using tools like
git,npm,dotnet, and thea365CLI. These operations are intended for project setup, dependency resolution, and deployment to cloud platforms like Azure, AWS, or GCP. - Indirect Prompt Injection Surface: The skill generates code that ingests user display names and interpolates them into agent instructions. While this creates a potential surface for indirect prompt injection, the skill includes explicit mitigation logic in the generated templates (e.g., the
_sanitize_display_namefunction in Python and .NET) to strip control characters and limit the length of the ingested text. - Environment Configuration: The skill updates environment configuration files (such as
.envandappsettings.json) with project-specific metadata fetched from the local environment. It follows industry best practices for managing these configurations, including reminders about secret management and recommending the use of secure secret stores for production environments.
Audit Metadata