The Agent Skills Directory

[SAFE]: The skill is purely instructional, providing templates and best practices for human intervention during agent execution. It contains no executable code or autonomous actions outside of documentation samples.
[DATA_EXFILTRATION]: The documentation describes patterns for collecting information such as passwords or API keys. This is presented as a legitimate use case for human-in-the-loop interaction where the user explicitly provides information to the agent. No evidence of unauthorized transmission to external domains was found.
[COMMAND_EXECUTION]: Code examples utilize standard Python functions and framework-specific hooks. There are no patterns indicative of arbitrary command execution or shell injection.
[EXTERNAL_DOWNLOADS]: The skill references the autogen library, which is an established and well-known framework. No untrusted remote scripts or packages are referenced or downloaded.

ag2-hitl