HasBrains Agent Kit — Production Claude Code Skills, Agents & Commands

The battle-tested Claude Code kit behind HasBrainsAI — agent skills, subagents, and slash commands for building production multi-agent systems. Every piece of this kit is used daily to ship real agent pipelines, not generated in bulk.

Works with Claude Code, and skills follow the portable Agent Skills standard used by Codex, Cursor, and Gemini CLI.

Install (one command)

/plugin marketplace add endorphin-ai/hasbrains-agent-kit

Then install only what you need:

/plugin install agent-engineering@hasbrains-agent-kit
/plugin install context-engineering@hasbrains-agent-kit
/plugin install content-creation@hasbrains-agent-kit
/plugin install teaching@hasbrains-agent-kit
/plugin install web-dev@hasbrains-agent-kit

Install a single skill

Don't need a whole plugin? Every skill is a self-contained folder following the portable Agent Skills standard, so you can cherry-pick one with the skills CLI:

npx skills add endorphin-ai/hasbrains-agent-kit --full-depth

It lists every skill in the kit and lets you pick the ones you want. To skip the picker, name the skill directly:

npx skills add endorphin-ai/hasbrains-agent-kit --full-depth --skill <skill-name>

Skills install into the current project's .claude/skills/ by default; add -g to install globally to ~/.claude/skills/. This also works for Cursor, Codex, and other agents that support the skills standard.

Or skip tooling entirely — a skill is just a folder:

cp -r teaching/skills/<skill-name> ~/.claude/skills/

What's inside

Plugin	What it does
`agent-engineering`	Engineering discipline for AI agent systems. Four always-on principles, test-driven development, verification-before-completion, system-architecture playbooks, a scope-disciplined product manager (PRD → roadmap → acceptance gate), and docs/-native project management.
`context-engineering`	Context engineering for AI agents. AI-optimized second-brain skills, durable pipeline state for multi-agent handoffs, expert prompt building.
`content-creation`	Content engine for technical creators. Social posts and video scripts engineered around named copywriting frameworks, scroll-stopping hooks, and retention mechanics.
`teaching`	Interactive AI teacher. `/repo-to-course` turns any codebase into a structured tutorial course — markdown, GitHub Pages site, or interactive HTML.
`web-dev`	Multi-agent web dev pipeline. UX research and design critique around the build (personas, storyboards, heuristics, measured design forensics), plus the designer ↔ frontend contract: freeze approved mockups into a canonical oracle, port them 1:1 on any stack, verify with screenshot-based design-diff gates.

Using the prompts

Raw prompt library lives in prompts/ — standalone prompts that don't map to a skill or command. Prompts aren't installed, they're plain markdown; two ways to use one:

Paste it — copy the file's content straight into your conversation.
Make it a slash command — copy the file into ~/.claude/commands/ (or your project's .claude/commands/) and it becomes invocable by filename, e.g. my-prompt.md → /my-prompt.

Using the output styles

Output styles live in output-styles/ — they change how Claude Code writes, not what it does. Copy one into your styles folder and select it:

cp output-styles/asd-ste100.md ~/.claude/output-styles/

Then run /output-style and pick it, or /output-style asd-ste100. Drop the file into a project's .claude/output-styles/ instead to scope it to that repo.

Style	What it does
`asd-ste100`	Concise, direct, technical writing based on ASD-STE100 Simplified Technical English. Result first, one idea per sentence, no filler or marketing language. Keeps coding instructions intact.

Using the utils

Utils live in utils/ — standalone tools for your Claude Code setup, not plugins. Each has its own installer:

git clone --depth 1 https://github.com/endorphin-ai/hasbrains-agent-kit.git
cd hasbrains-agent-kit/utils/statusline-kit && ./install.sh

Util	What it does
`statusline-kit`	Live status line: the active `/command` or skill run with a timer, per-subagent context bar, model, tokens, tool calls and estimated cost (priced per message from one editable `prices.json`), plus the session cost. One-command install, safe to re-run.

Design principles

This kit follows a strict separation of concerns:

Agents encode behavior and procedure only
Skills hold reusable know-how
Project-specific facts (stack, paths, commands, naming) stay in your project's own config — never baked into agents or skills

That's why these plugins drop into any codebase without modification.

Why trust this

Skills execute code in your environment — you should only install from sources you trust. This kit is maintained by Andrew Novykov, builder behind HasBrainsAI. Every skill and agent here is used in production on real projects, and most have a walkthrough video on the channel showing exactly how they work and what's inside.

Watch it in action

Deep dives on the architecture behind this kit — multi-agent orchestration, spec-driven development, agentic engineering for senior engineers:

youtube.com/@HasBrainsAI

License

MIT — see LICENSE.

Made by HasBrains — https://hasbrains.com/