🛡️ livey-secure-ai-workflow
_{Red-team inspired AI workflows for code & security review}

Universal AI-powered toolkit for terminal-first developers, hackers, and automation power users

Universal AI toolkit for terminal power users • LLM-agnostic • UAEP-powered • Privacy-first • SSH + automation

🔧 Overview

LiveY-Codex-Toolkit is an AI-augmented, terminal-native development toolkit that unifies:

Code review
Commit message generation
Patch suggestion + fix flow
Red-team static analysis
UAEP protocol execution
Structured agents for Laravel, security, Rust, Python and more
Git + SSH automation
Local AI support (Ollama, llama.cpp, local endpoints)

It is fully LLM-agnostic, supports any model, and remains privacy-first by design.

Built for developers who live in the terminal and run their workflows like a command center.

⚡ Why This Toolkit Exists

Because modern development deserves tools that:

write commit messages for you
analyze your code instantly
provide structured refactor plans
observe your entire repository safely
execute AI instructions consistently
work offline with local models
integrate into your git+SSH workflow
respect privacy and security boundaries

This toolkit tries to make AI feel like a real teammate — reliable, structured, predictable.

🧩 Features

🧠 AI-Augmented Commands

Command	Description
cx-uaep	Loads Universal Agent Execution Protocol (production, staging, special modes).
cx-review	Senior-level code review + security reasoning + structure evaluation.
cx-commit	Generates Conventional Commit messages automatically.
cx-fix	Debugs failing tests using AI-guided reasoning.
cx-security	Performs red-team static analysis on multi-language codebases.

🌐 LLM Compatibility

LiveY Codex Toolkit is fully LLM-agnostic.
Every command (cx-review, cx-pr, cx-fix, cx-security, UAEP modes, templates) works with:

✔ Cloud Models

Claude 3.5 / 3 / 2 (Anthropic)
GPT-4o / GPT-4.1 / GPT-3.5 (OpenAI)
Mistral Large / Medium / Small
DeepSeek-Coder / DeepSeek-V3
Qwen 2.5 Series
Gemini (via OpenAI-compatible gateways)

✔ Local Models

Ollama (Llama3, Qwen, Mistral, DeepSeek, Phi-3…)
LM Studio
llama.cpp servers
Any OpenAI-compatible local API (/v1/chat/completions)

✔ Custom Engines

Droid CLI
OpenRouter
litellm bridges
Self-hosted inference endpoints

🔐 API keys & safety

Add your provider key as a GitHub secret LLM_API_KEY; never commit keys to the repo.
Workflows auto-skip if LLM_API_KEY is missing and LLM_PROVIDER is not set to local.
To avoid egress, set LLM_PROVIDER=local and LLM_ENDPOINT=http://localhost:11434/api/generate.
Rotate keys regularly and scope them to least privilege.

Why it works

The toolkit uses:

codex as a universal abstraction layer
UAEP for structured, predictable responses
openai-compatible HTTP schema
explicit model profiles via TOML

This makes the toolkit portable, engine-independent, and future-proof.

🎭 Agent Profiles

Load with UAEP:

cx-uaep production --profile claude cx-uaep production --profile qwen cx-uaep staging --profile local

⟁🔄 Engine Switching — Runtime Model Selector

Switch engines dynamically:

cx-engine claude claude-3.5-sonnet cx-engine ollama llama3 cx-engine qwen qwen2.5 cx-engine droid neon-16k

Activate the new engine: codex --profile current cx-review

⟁🤖 Droid CLI Support — Alternate LLM Engine

Droid CLI can serve as a complete backend engine:

[mode.droid] provider = "openai-compatible" api_url = "http://localhost:8000/v1" model = "neon-16k"

Use it: codex --profile droid cx-review

🟣 Some Available CLI

🟣 VISUAL DIAGRAM

🟣 cx-engine-ui — TUI Selector

Create: bin/cx-engine-ui

#!/usr/bin/env bash

CY="\033[1;36m"; MG="\033[1;35m"; GR="\033[1;32m"; RS="\033[0m"

echo -e "${CY}⟁ LiveY Codex Toolkit — Engine Selector UI${RS}" echo -e "${MG}-------------------------------------------${RS}" echo "" echo -e "Select an engine:" echo "" echo -e " [1] Claude (claude-3.5-sonnet)" echo -e " [2] Ollama (llama3)" echo -e " [3] Qwen (qwen2.5)" echo -e " [4] Droid CLI (neon-16k)" echo -e " [5] Cancel" echo "" read -rp "Choice: " c

case "$c" in

engine="claude"; model="claude-3.5-sonnet" ;;
engine="ollama"; model="llama3" ;;
engine="qwen"; model="qwen2.5" ;;
engine="droid"; model="neon-16k" ;;
echo "Cancelled."; exit 0 ;; *) echo "Invalid option."; exit 1 ;; esac

echo "" echo -e "${CY}Setting engine → ${GR}${engine} (${model})${RS}" cx-engine "$engine" "$model"

echo "" echo -e "${GR}✔ Engine updated. Use:${RS} codex --profile current cx-review"

Make it executable: chmod +x bin/cx-engine-ui

🌐 LLM Compatibility

LiveY Codex Toolkit is fully LLM-agnostic.
Every command (cx-review, cx-pr, cx-fix, cx-security, UAEP modes, templates) works with:

✔ Cloud Models

Claude 3.5 / 3 / 2 (Anthropic)
GPT-4o / GPT-4.1 / GPT-3.5 (OpenAI)
Mistral Large / Medium / Small
DeepSeek-Coder / DeepSeek-V3
Qwen 2.5 Series
Gemini (via OpenAI-compatible gateways)

✔ Local Models

Ollama (Llama3, Qwen, Mistral, DeepSeek, Phi-3…)
LM Studio
llama.cpp servers
Any OpenAI-compatible local API (/v1/chat/completions)

✔ Custom Engines

Droid CLI
OpenRouter
litellm bridges
Self-hosted inference endpoints

Why it works

The toolkit uses:

codex as a universal abstraction layer
UAEP for structured, predictable responses
openai-compatible HTTP schema
explicit model profiles via TOML

This makes the toolkit portable, engine-independent, and future-proof.

🎭 Agent Profiles

Specialized LLM-optimized agents:

codex/prompts/agent.claude.md – Claude optimized codex/prompts/agent.qwen.md – Qwen optimized codex/prompts/agent.local.md – Local LLM optimized

Use with UAEP:

cx-uaep production --profile claude cx-uaep production --profile qwen cx-uaep staging --profile local

🔄 Engine Switching Switch LLM engine dynamically: cx-engine

Examples: cx-engine claude claude-3.5-sonnet cx-engine ollama llama3 cx-engine qwen qwen2.5 cx-engine droid neon-16k

Activate the profile: codex --profile current cx-review

🤖 Droid CLI Support

LiveY Codex Toolkit can use Droid CLI as an LLM backend via the included mode.droid profile:

🚀 What's New in v1.1

🎯 UAEP Integration

Universal Agent Execution Protocol (production + staging + special modes).

📁 Unified Prompt Architecture

All agents placed under:

codex/prompts/

🧪 New Execution Command

cx-uaep [mode]

💻 Better Local AI Support

Works with Ollama, LM Studio, and custom local endpoints.

📦 Installation

git clone [email protected]:ind4skylivey/livey-codex-toolkit.git cd livey-codex-toolkit mkdir -p ~/bin ln -s "$PWD/bin"/cx-* ~/bin/

#🔹Universal Installer ./install.sh

🔮 UAEP Protocol

UAEP standardizes AI execution using 5 strict response layers:

PLAN – architecture + reasoning

FILE-TREE – directory layout

GENERATION RULES – formatting, standards, safety

IMPLEMENTATION – full working code

TASK QUEUE – next actions

UAEP can run in: Production Mode

strict, safe, fully documented

ideal for real projects

Staging Mode

rapid prototyping

TODOs allowed

ideal for experiments

Special Modes

security_analysis

performance_max

ai_local_support

architecture_review

Load UAEP:

cx-uaep production cx-uaep staging cx-uaep production --mode security_analysis

🧠 Agent Profiles Profile Description agent.laravel.md Laravel-specialized developer agent agent.security.md Red-team static analysis agent agent.rust.md (WIP) Rust agent uaep/* Full UAEP protocol

Load via UAEP:

cx-uaep production --profile laravel

Direct usage:

codex --load codex/prompts/agent.laravel.md

🛠 Commands (Deep Documentation) cx-review

AI-powered repository inspection.

Examples:

cx-review cx-review working

cx-commit

Generates a clean Conventional Commit message.

cx-commit

cx-fix

AI-powered debugging of failing tests.

cx-fix cx-fix "npm test"

cx-security

Red-team static analysis:

cx-security cx-security src

📁 Project Structure

livey-codex-toolkit/ ├── bin/ │ ├── cx-uaep │ ├── cx-review │ ├── cx-commit │ ├── cx-fix │ └── cx-security │ ├── codex/ │ ├── prompts/ │ │ ├── uaep/ │ │ │ ├── uaep_core.md │ │ │ ├── uaep_production.md │ │ │ ├── uaep_staging.md │ │ │ ├── uaep_readme.md │ │ │ └── loader.md │ │ ├── agent.laravel.md │ │ ├── agent.security.md │ │ └── agent.rust.md (TODO) │ │ │ └── config/ (future) │ └── *.toml │ └── README.md

🔰 Philosophy & Safety

The toolkit never modifies files directly.

All changes are visible through diff previews.

You remain fully in control of commits, patches, and code writing.

UAEP ensures predictable, structured responses across any LLM.

Use the profile system for safety:

dev

analyze

git-ssh

livey_power_user (full-auto mode 😈)

🔄 Workflows Daily Dev Loop

cx-review cx-fix cx-commit git push

Security Sweep

cx-uaep production --mode security_analysis cx-security

Fast Prototyping

cx-uaep staging

Offline Mode

export CODEX_ENGINE_URL=http://localhost:11434 cx-review

🜁 LiveY Official Seal

⟁ ᛦ LIVEY · CODEX TOOLKIT ᛦ ⟁

Available soon as:

SVG (vector)

PNG (HQ)

Banner integration

🛣 Roadmap v1.2

install.sh

cx-pr

cx-changelog

execution_modes.toml

test suite

v2.0

Rust rewrite

Local AI fallback

Plugin system

TUI interface

v3.0+

multi-repo ops

CI/CD integrations

Neovim/VSCode bindings

template marketplace

🤝 Contributing

PRs welcome. Test on Linux. Document changes. Respect UAEP structure. 📜 License

MIT License. 🔗 Links

Repository: https://github.com/ind4skylivey/livey-codex-toolkit

UAEP Docs: codex/prompts/uaep/uaep_readme.md

Issues: https://github.com/ind4skylivey/livey-codex-toolkit/issues

Name		Name	Last commit message	Last commit date
Latest commit History 10 Commits
.github		.github
assets		assets
bin		bin
codex		codex
templates		templates
CHANGELOG.md		CHANGELOG.md
LICENSE		LICENSE
README.md		README.md
SECURITY.md		SECURITY.md
cx-changelog.sh		cx-changelog.sh
cx-local.sh		cx-local.sh
cx-pr.sh		cx-pr.sh
cx-template		cx-template
cx-update.sh		cx-update.sh
install.sh		install.sh

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Repository files navigation

🛡️ livey-secure-ai-workflow
_{Red-team inspired AI workflows for code & security review}

🔧 Overview

⚡ Why This Toolkit Exists

This toolkit tries to make AI feel like a real teammate — reliable, structured, predictable.