RayCodes GenericAgent — AI skill for Claude Code

🌱 GenericAgent Self-Evolving Web UI Filler

**Autonomous web form automation that learns from itself** — powered by [GenericAgent](https://github.com/lsdefine/GenericAgent), a local Ollama LLM (`gemma4-nothink`), and a Twitter/X lookalike dummy site.

---

📺 Demo Flow

First Run:   [Ollama: gemma4-nothink] → [Selenium fills form] → [Tweet posted] → [Skill saved]
Later Runs:  [Skill file found] → [exec()] → [Tweet posted] — ZERO LLM tokens used

---

🗂️ Project Structure

Generic Agent/
├── GenericAgent/            ← Cloned GenericAgent framework (the engine)
│   └── mykey.py             ← Ollama API config (you edit this)
├── dummy_site/
│   ├── index.html           ← Twitter/X lookalike UI
│   ├── server.py            ← Flask REST server (port 5050)
│   └── static/
│       └── style.css        ← Dark-mode glassmorphism stylesheet
├── skills/
│   └── fill_tweet.py        ← AUTO-GENERATED on first success (crystallized skill)
├── agent_filler.py          ← Main entry point / orchestrator
└── README.md                ← This file

---

🧰 Tech Stack

Layer	Technology	Why
Agent Framework	GenericAgent v1.0	Self-evolving skill tree, 6x fewer tokens than alternatives
LLM	gemma4-nothink:latest via Ollama	Local, fast (no CoT thinking tokens), privacy-preserving
Ollama API	OpenAI-compat /v1/chat/completions	Drop-in format, works with GenericAgent's oai_* config
Browser Automation	Selenium + webdriver-manager	Headful Chrome control, auto-installs correct ChromeDriver
Dummy Site	Flask + Vanilla HTML/CSS	Zero framework overhead, runs instantly
Styling	Inter font, dark glassmorphism	Twitter/X visual fidelity

---

🔑 Code Walkthrough

`GenericAgent/mykey.py` — LLM Configuration

oai_ollama_config = {
    'apikey':   'ollama'