CodeBOT

CodeBOT is a semi-autonomous AI development team for human-controlled workflows. One prompt becomes an Orchestrator-led plan for your approval-then four specialist roles execute the full implementation and report every change, with clear scope, handoffs, and verification. Built for GitHub Copilot App and Codex.

What’s Included in CodeBOT

The canonical profiles define five roles:

Role	Responsibilities
orchestrator	Clarifies scope, plans and directs tasks, coordinates specialists, checks approvals, verifies evidence, and reports results. Does not edit files or run shell commands.
data-analyst-manager	Analyzes requirements, data, risks, evidence, and recommendations.
application-uiux-designer-manager	Defines user requirements, UI/UX flows, visual direction, and design acceptance criteria.
software-engineer	Investigates issues, implements approved changes, debugs and tests, and reports technical evidence.
prompt-engineer	Prepares prompts and validated output contracts for handoff to specialists.

The profiles are available in profiles/canonical/.

How to Use CodeBOT

For now, use this repository as a human-readable reference for workflows and agent profiles. CodeBOT does not yet have a supported installer or runtime integration.

1. Clone or download the repository, then read the canonical profiles.
2. To experiment with an agent runtime, manually adapt the profiles to a format supported by that runtime. Do not assume that candidate adapter documents represent a working integration.
3. Follow the documented workflow: define the scope and acceptance criteria; assign tasks to the appropriate specialists; obtain explicit approval before actions that require it; return specialist results to the Orchestrator; and verify changes and evidence before reporting completion.

Example CodeBOT usage:

@orchestrator Redesign my portfolio website using a Soft Brutalist Editorial Tech design with the color codes DF301C, FF9100, FFF1D1, and 00B7CD. Do not change the website’s content. Execute.

Use @ to call the Orchestrator agent profile, which leads the specialist agents.

Recommended Usage in GitHub Copilot App

For GitHub Copilot App, it is recommended to invoke CodeBOT through the dynamic workflows mechanism. For example:

@orchestrator use CodeBOT dynamic workflows to redesign my portfolio website using a Soft Brutalist Editorial Tech design with the color codes DF301C, FF9100, FFF1D1, and 00B7CD. Do not change the website's content. Execute.

Using the phrase "use dynamic workflows to..." is strongly recommended to help ensure reliable orchestration and proper workflow execution.

During the code implementation phase, Autopilot mode is highly recommended. This helps prevent the orchestrator from falling back to a mode in which it may lack the necessary write tools, potentially disrupting implementation or causing the workflow to drift away from the intended task.

Supported Platforms (Current)

CodeBOT currently supports only the following agent platforms:

Platform	Status	Notes
GitHub Copilot App	Supported	Agent profiles must be placed/adapted according to the app's profile format.
Codex	Supported	Agent profiles and plugin-based skills are supported (see installation notes).

Other agent runtimes, CLIs, or chat integrations are not supported at this time. Do not assume that candidate adapter documents or third-party instructions establish compatibility with CodeBOT.

Installation

CodeBOT Repository

Run the following command to install the CodeBOT agent profiles automatically:

For macOS / Linux: Open Terminal. For Windows: Open Git Bash.

Copy the command below and press Enter:

for Terminal/Git Bash/WSL

curl -sSL https://raw.githubusercontent.com/ivanalvarizjr/CodeBOT/main/install.sh | bash

for powershell

Github Copilot App

Invoke-RestMethod -Uri "https://raw.githubusercontent.com/ivanalvarizjr/CodeBOT/main/install.sh" -OutFile "install.sh"; & "C:\Program Files\Git\bin\bash.exe" ./install.sh 1; Remove-Item "install.sh"

Codex

Invoke-RestMethod -Uri "https://raw.githubusercontent.com/ivanalvarizjr/CodeBOT/main/install.sh" -OutFile "install.sh"; & "C:\Program Files\Git\bin\bash.exe" ./install.sh 2; Remove-Item "install.sh"

Uninstalling CodeBOT

To uninstall CodeBOT from your local environment, run the appropriate command for the integration you previously installed.

GitHub Copilot App

Run the following command to uninstall the CodeBOT integration for GitHub Copilot App:

Invoke-RestMethod -Uri "https://raw.githubusercontent.com/ivanalvarizjr/CodeBOT/main/uninstall.sh" -OutFile "uninstall.sh"; & "C:\Program Files\Git\bin\bash.exe" ./uninstall.sh 1; Remove-Item "uninstall.sh"

Codex

Run the following command to uninstall the CodeBOT integration for Codex:

Invoke-RestMethod -Uri "https://raw.githubusercontent.com/ivanalvarizjr/CodeBOT/main/uninstall.sh" -OutFile "uninstall.sh"; & "C:\Program Files\Git\bin\bash.exe" ./uninstall.sh 2; Remove-Item "uninstall.sh"

Important Notes

• Select the command that matches your installed CodeBOT integration.
• Ensure that bash is available in your environment before executing the command.
• Review the uninstall script before running it to understand which local files and configurations it may remove.
• These commands download and execute the uninstall script directly from the CodeBOT repository. Make sure you trust the source before proceeding.

Required Skills and Optional Supporting Skills

Graphify is required for work involving project structure. Agent Reach is required for internet research, web/code searches, URLs, and online platforms. Taste Skill is required as a core skill for application design. Caveman, described below, is an optional supporting skill and is not a general prerequisite.

Graphify - Required

Graphify builds a searchable knowledge graph from project code and documentation. Its official PyPI package name is graphifyy (with two y characters); the command name is graphify.

Install the prerequisites for your operating system, then install Graphify:

Windows (PowerShell):

winget install astral-sh.uv
uv tool install graphifyy
graphify install

macOS (Homebrew):

brew install python@3.12 uv
uv tool install graphifyy
graphify install

Linux (Ubuntu/Debian):

sudo apt install python3.12 python3-pip pipx
uv tool install graphifyy
graphify install

For an alternative way to install uv on Ubuntu/Debian, Graphify also documents curl -LsSf https://astral.sh/uv/install.sh | sh; afterward, run uv tool install graphifyy. For a project-specific Codex installation, use:

graphify install --project --platform codex

Graphify also provides instructions for GitHub Copilot CLI and VS Code Copilot Chat. Those instructions do not establish compatibility with CodeBOT or GitHub Copilot App. See the official Graphify installation guide.

Agent Reach - Required for Internet Research

Agent Reach is a router and readiness checker for web, GitHub, RSS, video, and other platform tools. Use this skill for online research, web/code searches, URLs, and supported platforms. Agent Reach is not a command wrapper: once a channel is available, the agent uses the appropriate upstream tool.

Python 3.10 or later is required. The upstream guide recommends pipx; the package is installed directly from the official repository:

Windows (PowerShell; official venv method):

py -3 -m venv "$env:USERPROFILE\.agent-reach-venv"
& "$env:USERPROFILE\.agent-reach-venv\Scripts\Activate.ps1"
python -m pip install https://github.com/Panniantong/agent-reach/archive/main.zip
agent-reach install --env=auto

macOS and Linux (pipx):

pipx install https://github.com/Panniantong/agent-reach/archive/main.zip
agent-reach install --env=auto

agent-reach install --env=auto only checks readiness and does not modify the system. After installation, review the available channels:

agent-reach doctor --json

Installing system dependencies or optional channels requires explicit approval. After the user approves, an example for the OpenCLI channel is:

agent-reach install --env=auto --system --channels=opencli

Select only the channels you need; --channels=all may install many tools and modify system configuration. If Homebrew Python blocks a direct installation (PEP 668), use pipx or a virtual environment. On Windows, if python3 opens the Microsoft Store, use the Python Launcher py -3 as shown above. See the official Agent Reach installation guide and update guide.

Taste Skill - Required

Taste Skill provides design-taste-frontend for interface design work. The official Agent Skills installer is available on Windows, macOS, and Linux; Node.js and npm are required:

npx skills add https://github.com/Leonxlnx/taste-skill --skill "design-taste-frontend"

The skill scripts require Node.js 22 or later, as well as Chrome, Edge, or Chromium. For Codex, the repository also documents plugin installation:

codex plugin marketplace add Leonxlnx/taste-skill
codex plugin add taste-skill@taste-skill

Start a new Codex session after installing the plugin. See the official Taste Skill installation instructions.

Caveman - Optional

Caveman provides agent skills and tools, including a skill for concise responses. Its official interactive installer supports Windows, macOS, and Linux. Using npx requires Node.js 22.13 or later:

npx @caveman-ai/cli@latest

The repository also provides an installer for version v4.0.1:

Windows (PowerShell 5.1+):

irm https://raw.githubusercontent.com/JuliusBrussee/caveman/v4.0.1/install.ps1 | iex

macOS and Linux:

curl -fsSL https://raw.githubusercontent.com/JuliusBrussee/caveman/v4.0.1/install.sh | bash

These commands execute a downloaded installer script. Review the official installation instructions before running the script if you want to inspect it first. Check the Caveman repository for the list of supported agents and the latest configuration options.

Repository Guide

• profiles/canonical/ - canonical role profiles.
• [languange.md] - the language setting for agents. Indonesian and English are currently supported. Simply place this file in the workspace directory where the agent operates.

NOTES

For Codex Ensure that multi-agent support is enabled. If you are using an older version of Codex and agents are not detected, add the following configuration to your ~/.codex/config.toml file, or launch Codex with the --enable multi_agent flag:

[features]
multi_agent = true