agentleFS
JamalMohafilJamalMohafil/claude-skills/agent-starter-kit/SKILL.md

agent-starter-kit

Skill27 starsChanged 54 days agoMITView on GitHub

One-command setup for a new AI coding agent. Installs the essential starter skills (find-skills, agent-browser, karpathy-guidelines, ui-ux-pro-max, caveman, arabic-design, design-taste-frontend, remotion-best-practices) across 70+ agents, plus the core Claude Code plugins/MCPs (frontend-design, superpowers, code-review, playwright, github, claude-md-management). Use when the user is setting up a new project or a fresh agent and asks for "the skills/MCPs I should start with", a "starter pack", or "set up my agent".

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    If an agent loads this skill to set up an environment: run install.sh with the right --agent for the current agent.

    line 55

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Agent Starter Kit

The skills, plugins, and MCP servers worth installing the moment you open a new agent — bundled into one install. Instead of hunting them down one by one, run the installer and the agent starts with real capabilities: it can find and install more skills on demand, drive a browser, reason with guardrails, design clean UI (Arabic included), review its own code, and automate the browser + GitHub.

What it installs

Tier 1 — Skills (work on 70+ agents via npx skills: Claude Code, Codex, Cursor, …):

SkillWhat it gives the agent
find-skillsDiscovers and installs any other skill on demand — the meta-skill
agent-browserA real browser the agent controls itself (open, click, fill, verify)
karpathy-guidelinesBehavior rules: no silent assumptions, minimal solutions, verify before acting
ui-ux-pro-maxDesign-system intelligence — styles, color palettes, font pairings
cavemanBlunt, no-fluff answers when you want signal over politeness
arabic-designClean Arabic in AI-generated designs (fixes the letter-spacing/RTL traps)
design-taste-frontendDesign taste for landing pages, portfolios and redesigns — stops templated, generic UI (tasteskill)
remotion-best-practicesVideo in React with Remotion — the official skill, a router carrying 11 nested references (create · markup · captions · render · studio · maps · …)

Tier 2 — Plugins / MCPs (Claude Code only):

Plugin / MCPWhat it gives Claude Code
frontend-designAnthropic's design guidance for polished, non-generic UI
superpowersTDD + subagent orchestration workflow (brainstorm → design → build → review)
code-review/code-review on any branch/PR — scored, actionable review
playwright (MCP)Browser automation & end-to-end testing
github (MCP)Read/manage GitHub repos, PRs, and issues from the agent
claude-md-managementKeeps your CLAUDE.md files organized and current

How to run it

From the skill directory:

bash install.sh                 # Claude Code — installs everything
bash install.sh --agent cursor  # Cursor — installs the Tier 1 skills
bash install.sh --agent codex   # Codex  — installs the Tier 1 skills
bash install.sh --agent '*'     # every detected agent
bash install.sh --skills-only   # skip the Claude-Code-only plugins/MCPs

The script is best-effort and safe to re-run: each item installs independently, and anything it can't install automatically prints the exact manual command to finish it.

When invoked as a skill

If an agent loads this skill to set up an environment: run install.sh with the right --agent for the current agent. Then confirm what installed and surface any manual fallback lines the script printed. In Claude Code, remind the user to restart (or run /plugin) so the new plugins load.

Honest scope note

  • Tier 1 skills are genuinely cross-agent — npx skills targets 70+ agents.
  • design-taste-frontend is scoped by its own authors to landing pages, portfolios and redesigns — not dashboards, data tables, or multi-step product UI. It is also large (~87KB), so it costs real context when it fires. If a project's UI is a dashboard, or it already has a locked visual system, leave it out of the loop rather than letting it argue with the system you already decided on.
  • remotion-best-practices is one entry, not twelve. The official repo ships 12 skills, but the router installs with all 11 others nested inside it and every relative link resolving (verified). Installing the single router keeps the agent's skill list clean and loses nothing — do not "fix" this by adding the siblings as separate entries.
  • Tier 2 is Claude Code only. Codex/Cursor have no "plugin" equivalent. Their two MCP items (playwright, github) can still be added to Codex/Cursor through that agent's own MCP config — the script says so when you pass a non-Claude --agent.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.