conditional-chat-reminder
Pass
Audited by Gen Agent Trust Hub on Aug 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [Indirect Prompt Injection Surface]: The skill is designed to ingest and analyze external chat messages from Microsoft Teams to determine if a reminder should be suppressed. While the instructions explicitly state to treat chat content as 'untrusted data' and not as instructions, this remains a potential surface where adversarial text in a chat could attempt to influence the agent's logic during the evaluation phase. The skill mitigates this by using frozen baseline markers and specific update criteria.
- [Automated Command Execution]: The skill uses the
m_create_automationtool to schedule background tasks that involve reading chat history and posting messages. This is the core functionality of the skill, but users should be aware that it creates persistent automated actions that execute at a future time based on the state of external data.
Audit Metadata