Install etyb in
one command.

Two doors, one team — pick by the agent you're on. Claude Code → install the plugin. It ships the skill plus the five model-tiered sub-agents and the advisory hooks — the whole team, not just the knowledge. Any other agent → one npx command installs the same skill on every agentskills.io-compliant platform. Nothing to download or clone on either path.

/plugin marketplace add e-t-y-b/etyb-skills
/plugin install etyb@etyb-skills

$ npx skills add e-t-y-b/etyb-skills

On the npx path, the installer prompts you to choose which agents to wire up (defaults to the one you ran it from), what scope (project or global), and whether to symlink or copy. Symlink mode is the default — one source of truth in ~/.skills/, then linked into every agent's skills directory. Means one update propagates everywhere.

One skill — no per-tier choice, no peer skills competing at trigger time. Everything below ships in the box and is read from disk at runtime, on any harness.

The Claude Code plugin adds sub-agents and hooks on top — the full experience:

Native Claude Code plugin install — nothing to download or clone, Claude Code fetches the repo itself. Same skill content as the npx route, plus two things the npx route doesn't get:

Five sub-agents (explorer, planner, reviewer, stack-researcher, cartographer) carry out the heavy work — exploration, planning, review, vendor research, memory — in their own dispatched contexts, so your session stays clean. Model-tiered: light work runs on the smallest model tier; extensive-reasoning work inherits your own session model, which is always the ceiling.

Five advisory hooks fire outside the LLM — no test file before an edit, no review evidence before a commit, merging without a green test run. They surface a warning; they never block. Discipline is still primarily the model following instructions, with these as a visibility layer on top.

/plugin marketplace add e-t-y-b/etyb-skills

/plugin install etyb@etyb-skills

Does the npx install give me the sub-agents on Claude Code? No. npx skills add copies only the skill folder — the agent definitions and hooks live in the plugin manifest, so they arrive only via /plugin install. A skill-only install still runs /etyb with the full knowledge tree, but its heavy work runs instruction-guided instead of through the tuned, tool-scoped, model-tiered sub-agents. If you installed via npx first, remove it (npx skills remove etyb) and install the plugin — running both makes /etyb discoverable twice.

One command. Works on every agent that follows the agentskills.io spec — OpenAI Codex, Google Antigravity, Cursor, Cline, Gemini CLI, OpenCode, Warp, Amp, Deep Agents, Dexto, Firebender, Kimi Code CLI, and more. It runs on Claude Code too, but installs the skill only — on Claude Code, prefer the plugin so you also get the sub-agents and hooks.

$ npx skills add e-t-y-b/etyb-skills

There's no install script anymore — clone the repo and copy the skill folder directly into wherever your harness reads skills from. You get a local working copy you can fork, customize, or pin to a specific commit. Best for forks, air-gapped environments, or anyone who wants to read the source before installing.

$ git clone https://github.com/e-t-y-b/etyb-skills.git
$ mkdir -p .claude/skills
$ cp -r etyb-skills/skills/etyb .claude/skills/etyb

Swap the target directory per harness — .agents/skills/ for Codex, Antigravity, and Trae; .kiro/skills/ for Kiro; .cursor/skills/ for Cursor. The folder name must stay etyb, matching the name: in its SKILL.md frontmatter. Copy any stacks/<vendor>/ folders you want alongside it — they're read from disk the same way.

etyb ships a release roughly weekly — currency-stamped Stack updates and protocol refinements. Re-running update is the way to keep vendor knowledge fresh.

$ npx skills update etyb

$ npx skills update

Claude Code plugin: update from the /plugin menu, or run claude plugin update etyb@etyb-skills — the marketplace tracks releases and re-wires the hooks and agents automatically. Manual clone: cd etyb-skills && git pull, then re-copy skills/etyb to your target directory.

v4 ships as one folder. Remove it via the CLI or by deleting the symlink — the source in ~/.skills/ gets cleaned up automatically if no other agent references it.

$ npx skills remove etyb

$ rm -rf .claude/skills/etyb

Claude Code plugin: /plugin uninstall etyb@etyb-skills.

v3 (mid-2026 and earlier) shipped 30 separate sibling skills — one per specialist, protocol, and vertical — instead of the single /etyb skill every release since v4.0.0 ships. If any of those are still on disk, they compete with /etyb at trigger time. Remove them, then install fresh — there's no automated migration script anymore, but the fix is two steps:

$ rm -rf .claude/skills/research-analyst .claude/skills/tdd-protocol .claude/skills/backend-architect
$ npx skills add e-t-y-b/etyb-skills

Swap .claude/skills/ for whichever discovery directory your harness uses (see Manual / git clone above for the full list). Plans, local settings, and anything under .etyb/ are untouched either way. v3.x's per-tier bundles and their install flags are long gone — every release since v4.0.0 ships everything in the one skill.

Every command the universal installer exposes. Run npx skills --help for the full reference straight from the binary.

The universal installer auto-detects 14 agents at install time. All of them are agentskills.io-compliant; any future agent that adopts the spec works without a code change here.

• Claude Code
• OpenAI Codex
• Google Antigravity
• Cursor
• Cline
• Gemini CLI
• OpenCode
• Warp
• Amp
• GitHub Copilot
• Deep Agents
• Dexto
• Firebender
• Kimi Code CLI

The npx skills flags you'll actually reach for. Skip this section unless you need to override default behavior — see npx skills --help for the exhaustive reference.

npx skills add e-t-y-b/etyb-skills -g

npx skills add e-t-y-b/etyb-skills --list

npx skills add e-t-y-b/etyb-skills --all

npx skills add e-t-y-b/etyb-skills -y

Still stuck? Open an issue on GitHub or check the FAQ.