The Hands-on AI Plugin
One plugin, one install, four categories of work. Bundles the AI Workflow Framework (methodology), the Knowledge Graph skill (modeling the work itself), the AI Registry (documentation tooling), and Agentic Coding (feature-spec writing) — everything you need to take an AI workflow from idea to documented production system. 1 agent + 15 skills.
These are plain-text Markdown files that work in any AI tool. Install the plugin on Claude, ChatGPT, or Claude Code for the full experience, or download individual skills as ZIPs for Gemini, M365 Copilot, Cursor, and other platforms.
Get the Plugin
Section titled “Get the Plugin”The same plugin installs on Claude, ChatGPT, and Claude Code — pick your tab. Full step-by-step instructions for every platform, with screenshots, are on the skills setup page.
Works in Claude Chat (web and app) and Cowork on paid plans. In the left sidebar: Customize → Plugins → + → Add marketplace → Add from a repository, type jamesgray-ai/handsonai-plugins, click Sync, then Install on the handsonai card. The framework-agent orchestrator runs in Cowork.
On the Free plan, or if your company blocks plugins, use the ZIPs tab instead.
Works on any paid ChatGPT plan (Plus, Pro, Business, Enterprise) and in Codex. In the left sidebar: Plugins → Add marketplace, type jamesgray-ai/handsonai-plugins, then Install on Hands-on AI. Use a skill with @analyze in Work or $analyze in Codex. Workspace admins can add the marketplace for everyone under Admin → Plugins → Add → Import marketplace.
Don’t see Add marketplace? Upload the ZIPs instead (Plugins → Skills → Create → Upload).
Claude Code:
/plugin marketplace add jamesgray-ai/handsonai-plugins/plugin install handsonai@handsonaiCodex CLI:
codex plugin marketplace add jamesgray-ai/handsonai-pluginscodex plugin add handsonai@handsonaiSee Using Plugins for updates, uninstalling, and troubleshooting in Claude Code.
Every skill is also a standalone ZIP for platforms without the plugin — Claude Free, ChatGPT Free/Go, Gemini Spark and Gemini Enterprise, M365 Copilot Cowork, Cursor, Gemini CLI, VS Code Copilot, and more. Download from the skill downloads table (direct links, one per skill) or the GitHub Releases page, then follow your platform’s steps on the skills setup page.
What’s Inside
Section titled “What’s Inside”The plugin is organized into three coherent toolkits. Use them independently or together.
AI Workflow Framework
Section titled “AI Workflow Framework”The seven-step methodology for going from a workflow idea to a deployed AI system: Analyze → Deconstruct → Design → Build → Test → Run → Improve. Run the full flow with the orchestrator agent, or invoke any individual step as a skill.
framework-agent
Section titled “framework-agent”Walks you through the full 7-step framework end-to-end. The agent calls the right skill at each step, gathers inputs from you between steps, and produces working artifacts (specs, prompts, agents) along the way.
When to use it: Use this when you have a workflow you want to apply AI to and want guided, end-to-end help. If you’d rather drive each step yourself, invoke the skills directly.
Skills
Section titled “Skills”| Skill | Slash command | What it does |
|---|---|---|
analyze | /handsonai:analyze | Audit your workflows (individual or organizational lens) to find where AI creates the most value |
deconstruct | /handsonai:deconstruct | Break a workflow into structured steps using the 6-question framework |
design | /handsonai:design | Design the AI workflow architecture and produce a Design Spec |
build | /handsonai:build | Generate platform-appropriate artifacts (prompts, skills, agents) from the approved spec |
test | /handsonai:test | Test workflow artifacts and evaluate output quality |
run | /handsonai:run | Generate a Run Guide for deploying and operating the workflow |
improve | /handsonai:improve | Evaluate a running workflow for quality and evolution opportunities |
Full documentation: The AI Workflow Framework walks through the methodology with worked examples for each step.
Knowledge Graph
Section titled “Knowledge Graph”Model the work itself — clients, offerings, processes, people, tools, policies — as a knowledge/ graph your AI maintains from now on, following the Open Knowledge Format. Distinct from the AI Registry below, which records what you build with AI; the two bundles sit side by side.
building-knowledge-graph
Section titled “building-knowledge-graph”Slash command: /handsonai:building-knowledge-graph
Builds your business knowledge graph: a ~30-minute guided build that looks at what is already in your folder, interviews you about the work a person, team, or company does, proposes the four to seven kinds of things it runs on and the sentences that connect them, and, on the single word build, writes the knowledge/ bundle in the Open Knowledge Format — SCHEMA.md, seeded concept pages, an overview, index and log — plus a standing-rules file and local ingest and lint skills, then runs the loop once with you. Say “Build my knowledge graph.” Needs a tool that writes files on your computer (Claude Code, Cowork, the ChatGPT desktop app). Re-running on an existing graph fills gaps and never rebuilds.
AI Registry
Section titled “AI Registry”Once you’ve built workflows worth tracking, you need a registry. These five skills maintain your AI Registry — a registry/ knowledge bundle at your workspace root, plus everything derived from it. The bundle holds typed concept nodes (Business, LineOfBusiness, Function, Process, Workflow, Note) as individual Markdown files, governed by a registry/SCHEMA.md producer profile that every skill reads before it writes — required frontmatter, enum values, link discipline, and banned fields all live there, so every skill’s writes and every maintenance pass’s lint checks agree on the same contract. The bundle is the source of truth; the REGISTRY.md dashboard, directory indexes, and GENERATED blocks are all derived from it — edit sources, regenerate, never hand-edit a derived view.
Skills
Section titled “Skills”scaffolding-registry
Section titled “scaffolding-registry”Slash command: /handsonai:scaffolding-registry
Stands up a brand-new AI Registry: a guided, ~30-minute interview that creates the registry/ bundle — SCHEMA.md, the typed directories, and a first real Business, Line of Business, Function, Process, and Workflow node for your actual work. Triggers when you ask to set up your registry, or automatically when another framework skill finds no registry/SCHEMA.md in the workspace. Re-running on an existing bundle fills gaps rather than re-scaffolding, and it also migrates legacy layouts (old per-workflow manifest files or flat requirements files) into the bundle. Hands off to indexing-registry for the workspace’s first maintenance pass and dashboard.
naming-workflows
Section titled “naming-workflows”Slash command: /handsonai:naming-workflows
Generates consistent, outcome-focused names and descriptions for business workflows, then records them as Workflow nodes in your registry bundle. Follows domain-specific naming patterns (Sales, Marketing, Product, Education, Operations) so your registry stays organized as it grows.
writing-workflow-sops
Section titled “writing-workflow-sops”Slash command: /handsonai:writing-workflow-sops
Writes Standard Operating Procedure documentation for workflows and saves them as Markdown files, recording the SOP link in the workflow’s registry entry. Selects a full or lightweight SOP template based on autonomy level (deterministic vs. guided/autonomous) and adapts for execution mode (Manual, Augmented, Automated).
writing-process-guides
Section titled “writing-process-guides”Slash command: /handsonai:writing-process-guides
Writes Business Process Guide documentation that explains the strategic context and rhythm of a complete business process — when to execute it, why it matters, and how its component workflows fit together. The strategic companion to tactical SOPs.
indexing-registry
Section titled “indexing-registry”Slash command: /handsonai:indexing-registry
Maintains the AI Registry bundle and its derived views. Lints your bundle against its SCHEMA, refreshes indexes and GENERATED views, and regenerates your REGISTRY.md and visual dashboard. Runs five jobs in a full pass: lint the bundle, refresh typed-directory indexes, regenerate GENERATED blocks (a Workflow’s Insights, a Function’s Owns list), regenerate the REGISTRY.md index (and the visual dashboard registry-dashboard.html on request), and append log.md for schema changes and migrations. The framework skills trigger it automatically at their closing steps; run it standalone to pick up assets created outside the framework or to repair a stale index.
Recommended workflow: Scaffold the bundle → Name workflows → Document procedure (SOP) → Connect workflows (Process Guide) → Refresh the index. Each skill works independently if you don’t need the full sequence.
Agentic Coding
Section titled “Agentic Coding”For turning ideas into shippable feature specs before you write code. Two skills — one for fuzzy ideas, one for production specs.
Skills
Section titled “Skills”writing-vision-briefs
Section titled “writing-vision-briefs”Slash command: /handsonai:writing-vision-briefs
Captures a fuzzy idea as a structured Vision Brief — the essence of an idea before formal requirements. Walks you through problem, users, vision, capabilities, and success criteria in plain language. Assesses scope and breaks bigger visions into epics and features. Feeds directly into writing-feature-prds.
writing-feature-prds
Section titled “writing-feature-prds”Slash command: /handsonai:writing-feature-prds
Creates a Product Requirements Document for one feature with user stories, acceptance criteria, and a linked GitHub issue. Stress-tests the PRD for edge cases and ambiguity before finalizing.
Full documentation: Agentic Coding shows the full feature-development workflow these skills plug into.
Looking for example components?
Section titled “Looking for example components?”Earlier versions of the marketplace shipped specialized agents for executive writing, HBR-style editorial review, AI news research, and meeting prep, plus several portable prompts. They’ve moved to the Example Gallery as study material — copy and customize them for your own workflows.
License
Section titled “License”MIT — fork, modify, and ship freely.