agent-toolkit

DiUS's shared toolkit of skills, commands, agents and hooks for AI coding agents.

agent-toolkit is a monorepo where DiUS consultants publish and reuse the working pieces they build for AI coding agents — day-to-day tools, not one-off prompts. It's a growing collection, installable as a whole or one component at a time, that stays tool-agnostic wherever possible and installs as a Claude Code plugin in one command.

---

Install

Any agent — npx skills (recommended)

Installs skills into any of ~55 supported agents (Claude Code, Cursor, Codex, OpenCode, Gemini, Copilot…) straight from GitHub — the most tool-agnostic route:

# interactive — detects your agents and prompts, installs the whole toolkit
npx skills add DiUS/agent-toolkit

# install a single skill only
npx skills add DiUS/agent-toolkit --skill codebase-discovery

# preview what would be installed
npx skills add DiUS/agent-toolkit --list

Claude Code (plugin — skills + commands + agents)

/plugin marketplace add DiUS/agent-toolkit
/plugin install agent-toolkit@dius-agent-toolkit

Local / development:

git clone https://github.com/DiUS/agent-toolkit.git
claude --plugin-dir /path/to/agent-toolkit

See docs/getting-started.md for per-tool notes (Cursor, Gemini CLI, and any other agent).

---

What's in the toolkit

Skill	Description
codebase-discovery	Reverse-engineers domain, architecture, business rules, workflows and a business glossary from an existing codebase into lean onboarding docs — ready for harness engineering / Spec Kit.

New skills, commands, agents and hooks are added over time — see CONTRIBUTING.md to add your own.

---

Component types

• Skills (skills/) — self-contained Markdown workflows (an orchestrator plus its own playbooks, references and templates). The most portable component; any capable agent can follow one.
• Commands (commands/) — single Markdown files with frontmatter that become slash commands in Claude Code and compatible agents.
• Agents (agents/) — Claude Code subagent definitions a skill can delegate work to (scoped to the minimum tools they need, read-only where they only inspect code).
• Hooks (hooks/) — optional, Claude-Code-specific automations. Not enabled by default; see hooks/README.md.

---

Repository structure

agent-toolkit/
├── .claude-plugin/
│   ├── marketplace.json          # /plugin marketplace add …
│   └── plugin.json               # registers ./skills, ./commands, and the bundled agents
├── skills/
│   └── <name>/                   # self-contained skill — travels as one unit
├── commands/                     # slash commands (README.md explains the convention)
├── agents/                       # Claude Code subagents a skill can delegate to
├── hooks/                        # optional, opt-in Claude Code hooks (not wired in)
├── docs/
│   └── getting-started.md        # per-tool install notes
├── scripts/validate.js           # verification gate (manifests, frontmatter, paths)
├── .github/workflows/validate.yml# CI: runs the gate on push/PR
├── AGENTS.md                      # working context for any agent contributing here
├── CLAUDE.md                      # Claude Code specifics (imports AGENTS.md)
├── CONTRIBUTING.md
├── README.md
├── LICENSE
└── .gitignore

---

Contributing / working on this repo

This repo is itself harness-ready: AGENTS.md gives any agent the context and guardrails to contribute (and CLAUDE.md adds Claude Code specifics), while CONTRIBUTING.md covers the mechanics of adding a skill, command, agent or hook. Before committing, run the verification gate — also enforced in CI:

node scripts/validate.js

License

MIT