The Agent Skills Directory

[Indirect Prompt Injection Surface]: The skill ingests and analyzes content from Microsoft Teams messages, which are external and untrusted inputs. This creates a surface for indirect prompt injection where instructions embedded in chat messages could attempt to influence the agent's summary or behavior. Ingestion points: Untrusted data enters the agent context through the workiq-ask tool when pulling messages from channels in Step 3 and searching for mentions in Step 5. Boundary markers: The instructions lack explicit boundary markers (such as XML tags or dedicated delimiters) to separate the untrusted message content from the core instructions, which may allow embedded text to be interpreted as agent commands. Capability inventory: The agent is granted access to the workiq-ask tool, which allows it to retrieve user profile information, list teams/channels, and read message history across the tenant. Sanitization: The skill does not describe specific sanitization or filtering steps to be applied to the message content before it is categorized and summarized, relying on the agent's internal safety guardrails.

channel-digest