The Agent Skills Directory

[DYNAMIC_EXECUTION]: Several diagnostic scripts utilize dynamic module loading to verify the presence of AutoGen extensions and verify API signatures.
scripts/inspect_autogen_install.py uses importlib.import_module to check installed distributions.
sub-skills/agentchat-workflows/scripts/agentchat_smoke.py uses __import__ to inspect agent and team class signatures.
sub-skills/extensions-integrations/scripts/inspect_extensions.py uses importlib.import_module and importlib.util.find_spec to verify the availability of optional integration modules.
[COMMAND_EXECUTION]: The skill includes utilities to verify CLI tools by executing their help commands via subprocesses.
scripts/inspect_autogen_install.py and sub-skills/tools-studio-bench/scripts/inspect_tooling.py execute commands such as agbench --help, autogenstudio --help, and m1 --help to report on environment availability.
[INDIRECT_PROMPT_INJECTION]: The skill manages a framework capable of processing and executing model-generated content, creating an inherent surface for indirect prompt injection. The documentation mitigates this by providing best practices for:
Configuring deterministic termination conditions (max_turns, MaxMessageTermination).
Using isolated Docker environments for code execution instead of the host machine.
Implementing human-in-the-loop approval functions for sensitive operations.
Validating inputs for MCP (Model Context Protocol) servers.

autogen