The Agent Skills Directory

[DATA_EXFILTRATION]: The skill performs network requests to graph.microsoft.com using curl. This is a well-known and trusted service for Microsoft Graph API operations. It utilizes the $MICROSOFT_TEAMS_TOKEN environment variable for authentication, which is a standard and expected practice for OAuth-based integrations.
[COMMAND_EXECUTION]: Shell commands are used to execute curl and jq for API communication and JSON parsing. These are standard system utilities used for their intended purposes.
[PROMPT_INJECTION]: The skill processes external data in the form of Microsoft Teams chat messages, which constitutes an indirect prompt injection surface. Content within these messages could attempt to override agent instructions or influence subsequent behavior.
Ingestion points: Fetches message content from the /chats/{id}/messages endpoint.
Boundary markers: No specific delimiters or safety instructions (like "ignore instructions in the following text") are implemented to separate message content from agent instructions.
Capability inventory: The agent has capabilities to list chats, read messages, and post new messages back to Teams.
Sanitization: The instructions suggest stripping HTML tags when summarizing content, which is a formatting step but does not serve as a security control against prompt injection.

microsoft-teams