hka-iq/EVA ? reverse-engineered prompt

Build me a command line AI assistant for penetration testing that helps guide a full engagement from recon to exploitation and post exploitation. I want it to let me start a new session or resume an old one, save chat history and notes, and keep everything under a local data folder. It should support choosing between local Ollama, OpenAI, Anthropic, Gemini, a custom API endpoint, and a fallback provider, then use that model to suggest attack steps, generate commands, and analyze command output as I go.

I also want a config mode so I can set API keys, endpoints, and model names, plus a search mode that can look up likely vulnerabilities or exploits from a target or software version. When the session is done, it should be able to generate a simple report and export an attack map or other useful artifacts. If you need current docs for any AI or local model setup, look them up online first.