SkyworkAI/DeepResearchAgent ? reverse-engineered prompt

Build me a Python app that acts like a deep research agent and can also be used for general task solving.

I want it to have a top level planner that can break a request into smaller steps, then hand work off to specialized agents and tools, and keep track of what happened along the way. It should be able to remember useful information between steps, trace its actions, and support versioned updates so the agent can improve over time without breaking everything. If a task needs outside capabilities, it should be able to call tools and interact with simple environments like files or other runtime state.

Please make it runnable with a clean example so I can try a tool calling agent right away, using a config file and an env file for the model key. Also set it up so I can swap models, prompts, memory, and tools through configuration instead of changing code. If you need to look up current docs online, feel free to do that.