ChallengeAI-CLI

Plugin, skill, and CLI development workspace for ChallengeAI agents.

This repository exists to build thin layers of reusable value for agent runtimes including Codex, Claude Code, and Clawhub/OpenClaw. It is not intended to be tied to Cursor or any IDE-specific execution model.

Purpose

This repo is the platform layer for:

• Agent plugins exposed through /plugins
• Reusable skills that can be invoked directly by a user or orchestrated by an agent
• Supporting CLI surfaces that make those skills operational outside an IDE
• Shared guidance for extending the ChallengeAI ecosystem consistently

The immediate direction is to expand these skills and plugin surfaces so they can support all 12 ChallengeAI-Web applications. The upstream application set is currently housed in orangeai and will be renamed to ChallengeAI-Web.

What This Repo Is Not

• Not an AWS landing zone
• Not an ECS provisioning template
• Not a Cursor-only workflow
• Not a manifest store for cloud infrastructure inventory

Older documentation in this repo described an agent-driven AWS provisioning flow. That is no longer the defining purpose of the repository.

Core Concepts

Plugin

A plugin is the packaged integration surface made available to an agent runtime. Plugins expose capabilities through runtime-specific mechanisms such as /plugins and can bundle one or more skills, prompts, helpers, and supporting metadata.

Skill

A skill is a narrow, reusable unit of agent behavior. Skills should be composable, easy to orchestrate, and focused on a concrete user outcome rather than broad domain ownership.

CLI

CLI tooling in this repo should provide automation, packaging, validation, or local developer workflows that support plugin and skill delivery across supported runtimes.

Supported Runtimes

• Codex
• Claude Code
• Clawhub/OpenClaw

The repo should prefer abstractions that transfer across runtimes cleanly. Runtime- specific behavior is acceptable when required, but it should be isolated and explicitly documented.

Install

npm

npm install -g @metaphase-tech/challengeai-cli

Homebrew

brew tap MetaPhase-Consulting/challengeai-cli https://github.com/MetaPhase-Consulting/ChallengeAI-CLI
brew install MetaPhase-Consulting/challengeai-cli/challengeai-cli

Run

challengeai

For full install, upgrade, uninstall, and runtime handoff guidance, see docs/reference/installing-challengeai-cli.md.

Documentation Map

• PLUGIN_ARCHITECTURE.md defines the platform model, packaging boundaries, and runtime expectations.
• APP_MAPPING.md captures how thin-value skills will map onto the 12 ChallengeAI-Web applications.
• docs/requirements/README.md indexes the requirements corpus used to plan, track, and validate plugin capability work.
• docs/iso42001/README.md indexes the ISO/IEC 42001 governance and audit artifacts for the plugin platform.
• docs/reference/README.md indexes reference assets including the suite “about” document and skill mapping.
• docs/reference/installing-challengeai-cli.md documents the current npm install flow, managed marketplace layout, and runtime handoff into Codex or Claude Code.
• Formula/challengeai-cli.rb is the Homebrew formula maintained alongside the npm package until it is extracted into a dedicated MetaPhase tap.
• .agent/build-guidelines.md defines the authoring guardrails for agents making changes in this repo.
• .agent/build-prompt.md is the implementation prompt for creating or extending plugin and skill work.
• .agent/teardown-prompt.md describes how to retire or remove plugin/skill work safely.
• .agent/manifest-template.md defines the work manifest shape for plugin and skill changes.

Working Model

1. Identify the target runtime or cross-runtime capability.
2. Determine whether the work belongs in a plugin, a skill, a CLI command, or a combination of the three.
3. Map the work to a ChallengeAI-Web app or a shared platform capability.
4. Implement the thin-value layer with minimal runtime-specific branching.
5. Record the work in a manifest if the task spans multiple assets or requires coordinated follow-up.

Repository Structure

.claude-plugin/            Claude Code plugin metadata
.codex-plugin/             Codex plugin metadata
.agent/                    Agent guidance and prompts for work in this repo
docs/                      Planning, requirements, compliance, and audit artifacts
docs/requirements/         Capability requirements, release gates, and design constraints
docs/iso42001/             ISO/IEC 42001 governance, evidence, and audit tracking
docs/reference/            Technical reference assets and suite identity docs
bin/                       Published ChallengeAI executable entrypoint
Formula/                   Homebrew formula definitions for ChallengeAI distribution
scripts/                   Packaging and release helper scripts
src/                       Installer, validation, and runtime-manager implementation
manifests/                 Optional work manifests for plugin/skill changes
skills/                    Cross-runtime skill definitions
PLUGIN_ARCHITECTURE.md     Platform architecture and packaging model
APP_MAPPING.md             ChallengeAI-Web app mapping and expansion plan

Naming Direction

• Use ChallengeAI-Web as the forward-looking product family name in new docs
• Reference orangeai only when needed to explain current upstream naming or repo history
• Keep plugin and skill names descriptive, runtime-neutral where possible, and scoped to the smallest reusable value layer