The Agent Skills Directory

[SAFE]: The skill provides legitimate documentation and code examples for the AG2 framework. It follows security best practices by recommending the use of environment variables and .env files for API key management rather than hardcoding credentials.
[COMMAND_EXECUTION]: The skill includes a diagnostic command for verifying the installation environment (python -c ...). This command is benign and intended for troubleshooting installation issues.
[INDIRECT_PROMPT_INJECTION]: The examples demonstrate processing user input via agent.ask(). While this represents a standard input surface for LLMs, the provided examples do not include any dangerous tools or system capabilities that could be exploited.
Ingestion points: agent.ask() and reply.ask() methods in hello_agent.py and multi_turn.py.
Boundary markers: Not implemented in these basic examples.
Capability inventory: No file system writes, network exfiltration, or shell execution capabilities are defined in the agent configurations.
Sanitization: Standard LLM processing without specific external sanitization logic.

ag2-quickstart