Agents & Skills
A focused toolkit of agents and skills that give your AI tool everything it needs to design, build, and document AI workflows. Plain-text Markdown — install the plugin on Claude, ChatGPT, or Claude Code, or download individual skills as ZIPs for Gemini, M365 Copilot, Cursor, and other tools.
🛠️ The Hands-on AI Plugin
Section titled “🛠️ The Hands-on AI Plugin”One plugin, one install. Bundles the AI Workflow Framework, the AI Registry skills, and the Agentic Coding feature-spec skills — everything you need to take an AI workflow from idea to documented production system.
| Where | How to install |
|---|---|
| Claude (Chat and Cowork, paid plans) | Customize → Plugins → + → Add marketplace → Add from a repository → jamesgray-ai/handsonai-plugins → Install |
| ChatGPT (paid plans) and Codex | Plugins → Add marketplace → jamesgray-ai/handsonai-plugins → Install Hands-on AI |
| Claude Code | /plugin marketplace add jamesgray-ai/handsonai-plugins then /plugin install handsonai@handsonai |
| Everything else | Download the skills as ZIPs |
→ See what’s included · Step-by-step setup for every platform · Using Plugins in Claude Code
What’s inside
Section titled “What’s inside”AI Workflow Framework — 1 agent + 7 skills
Section titled “AI Workflow Framework — 1 agent + 7 skills”The seven-step methodology for going from a workflow idea to a deployed AI system.
| Component | What it does |
|---|---|
framework-agent agent | Walks you through the full 7-step framework end-to-end |
analyze skill | Audit your workflows to find where AI creates the most value |
deconstruct skill | Break a workflow into structured steps using the 6-question framework |
design skill | Design the AI workflow architecture and produce a Design Spec |
build skill | Generate platform-appropriate artifacts from the approved spec |
test skill | Test workflow artifacts and evaluate output quality |
run skill | Do the first real run and leave a Run Card, run log, and review date |
improve skill | Evaluate a running workflow for quality and evolution opportunities |
Knowledge Graph — 1 skill
Section titled “Knowledge Graph — 1 skill”For modeling the work itself — clients, offerings, processes, people, tools, policies — as a graph the AI maintains.
| Component | What it does |
|---|---|
building-knowledge-graph skill | Build your knowledge/ knowledge graph — a guided interview finds the types and relationships of your work, then writes the schema, seeded pages, standing rules, and local ingest and lint skills |
AI Registry — 5 skills
Section titled “AI Registry — 5 skills”For documenting and operating an AI workflow registry once you’ve built workflows worth tracking.
| Component | What it does |
|---|---|
scaffolding-registry skill | Stand up your registry/ knowledge bundle — SCHEMA.md and real Business, Line of Business, Function, and Process nodes (Analyze adds the workflows) |
naming-workflows skill | Apply consistent naming conventions across your registry |
writing-workflow-sops skill | Author standard operating procedures for AI-assisted workflows |
writing-process-guides skill | Document multi-step processes for repeatable execution |
indexing-registry skill | Lint your registry bundle and regenerate REGISTRY.md and its visual dashboard — the derived views of your skills, agents, workflows, and apps |
Agentic Coding — 2 skills
Section titled “Agentic Coding — 2 skills”For turning ideas into shippable feature specs before you write code.
| Component | What it does |
|---|---|
writing-vision-briefs skill | Capture a fuzzy idea as a structured Vision Brief before writing a PRD |
writing-feature-prds skill | Create a PRD with user stories, acceptance criteria, and a GitHub issue |
🔗 The Multi-Agent Example Plugin
Section titled “🔗 The Multi-Agent Example Plugin”A separate, deliberately small plugin containing one complete multi-agent pipeline — four specialist agents that research, write, edit, and publish a business article, coordinated by Claude Code with quality gates and a human approval step.
It is a worked example, not a toolkit. Install it to see how automatic delegation, subagent handoffs, SubagentStop and PreToolUse hooks, and a human-in-the-loop gate fit together in something that actually runs — then adapt the shape to your own work.
/plugin marketplace add jamesgray-ai/handsonai-plugins/plugin install multi-agent-example@handsonai| Component | Type | Role |
|---|---|---|
ai-productivity-researcher | Agent | Finds case studies with quantified, sourced outcomes |
tech-executive-writer | Agent | Drafts the article from the research dossier |
hbr-editor | Agent | Produces the publication-ready revision plus an editorial memo |
hbr-publisher | Agent | Produces both deliverables and inspects its own output |
editing-hbr-articles | Skill | The editorial standards the editor loads |
subagent-gate.sh | Hook | Validates each stage’s artifacts before the agent may finish |
publish-gate.sh | Hook | Blocks publishing until a human approves |
/hbr-article | Command | Automatic delegation — Claude picks the specialists |
/hbr-article-strict | Command | Deterministic — the same pipeline, explicit sequence |
Why it is a separate plugin. It ships hooks, which change how Claude Code behaves in your projects. Installing the Hands-on AI plugin should never silently do that, so this is its own explicit opt-in. Both hooks stay completely inert unless a pipeline run is in progress.
Claude Code users also need Anthropic’s document skills for the Word output (/plugin install document-skills@anthropic-agent-skills). Cowork users need nothing — it is already included.
🎨 Example Gallery
Section titled “🎨 Example Gallery”Looking for an agent that writes for executives? A researcher that scans AI news daily? A skill for prepping meetings? Templates you can copy and customize live in the Example Gallery — 7 agents, 5 skills, and 3 portable prompts adapted from real workflows.
These are not bundled in the plugin because they’re scoped to specific personas, source lists, or industries. They’re better as study material than installable defaults.