The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends installing official and widely recognized libraries from Microsoft and Azure (e.g., azure-identity, msgraph-sdk, botbuilder-core) to interact with Teams and Graph APIs. This is a standard and safe practice for this domain.
[PROMPT_INJECTION]: The skill implements a message handler in bot.py that ingests untrusted data from Teams users (turn_context.activity.text). This represents a typical attack surface for indirect prompt injection in chat-based agents. However, the provided examples demonstrate benign processing (normalization and command matching) without exposing high-risk capabilities like shell execution or file system writes to the user input.
Ingestion points: bot.py -> on_message_activity (processes user text).
Boundary markers: None explicitly shown in the code snippets.
Capability inventory: The skill uses standard Bot Framework tools for messaging and Graph API for meeting/channel management. No dangerous functions (eval, exec) are used on external data.
Sanitization: Basic normalization (lowercasing and stripping) is used for input processing.
[COMMAND_EXECUTION]: The documentation provides standard shell commands for project setup, dependency installation, and environment variable configuration. These commands are transparent and intended for the developer to execute locally during development.

teams-api