maurorisonho/smol-autonomous-agent

Smol Autonomous Agent (AI Agents Course - Advanced Track)

This repository contains a production-ready Autonomous Code Agent (CodeAgent) built using Hugging Face's official smolagents library, developed for the Hugging Face AI Agents Course and Smol Course.

🛠️ Tools & Capabilities (@tool)

1. Model Security Auditor: Inspects Hub repositories for serialization risks (Pickle vulnerabilities), zero-code safetensors compliance, and documentation completeness.
2. Educational Benchmark Evaluator: Live validation of course passing thresholds across all Hugging Face educational spaces.
3. Sandbox Code Execution: Autonomous multi-step Python code generation and execution.

🚀 Quickstart

from smolagents import CodeAgent, InferenceClientModel
from agent import audit_hf_model_security, verify_course_benchmark

model = InferenceClientModel("Qwen/Qwen2.5-Coder-32B-Instruct")
agent = CodeAgent(tools=[audit_hf_model_security, verify_course_benchmark], model=model)
response = agent.run("Audit the security posture of maurorisonho/ppo-LunarLander-v2")
print(response)