The Agent Skills Directory

[SAFE]: The skill demonstrates security best practices by implementing a PathGuardian observer that monitors and blocks unauthorized access to system directories like /etc/ or /usr/ before they can be executed.
[COMMAND_EXECUTION]: The write_file tool provided in the safety demonstration is a mock implementation that does not perform actual filesystem operations, ensuring the example code is safe for evaluation and testing.
[DATA_EXFILTRATION]: No unauthorized network activity or data exfiltration patterns were detected. Monitoring state for token usage and repetition is maintained locally within the framework's event stream.
[PROMPT_INJECTION]: The agent prompts in the examples are purely instructional for demonstrating monitoring capabilities and do not contain directives to bypass safety filters or override agent constraints.

ag2-observers-and-alerts