agent-starter
Health Gecti
- License — License: MIT
- Description — Repository has a description
- Active repo — Last push 0 days ago
- Community trust — 78 GitHub stars
Code Gecti
- Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Gecti
- Permissions — No dangerous permissions requested
Bu listing icin henuz AI raporu yok.
Advanced Agentic framework: 40+ skills, meta-commands, TOON format (30-60% token savings), workflow automation & skill orchestration. npx create-agent-starter@latest
agent-starter
An opinionated multi-agent skill pack for Claude Code, Codex, and Cursor. Seven deep, handwritten skills for fintech, SaaS, mobile, copywriting, agent tooling, and TOON token savings.
No orchestration framework. No aspirational YAML. Just agent-native project files generated from one shared skill source.
What you get
7 top-level skills:
| Skill | Covers |
|---|---|
| stripe | Checkout, Payment Intents, subscriptions, Connect/marketplace, Terminal, Radar, Treasury, Issuing, webhooks. |
| supabase | Postgres + RLS, Auth with SSR cookies, Realtime, Storage, Edge Functions, pgvector. |
| plaid | Link flow, Auth, Transactions sync, Identity, Accounts, balances. |
| expo | EAS Build, EAS Update, Expo Router, React Native app patterns. |
| copywriting-frameworks | Headlines, landing pages, ads, emails, CTAs, AIDA, objections, proof placeholders, critiques. |
| anthropic | Anthropic Claude API plus Claude Code meta-tooling sub-skills. |
| toon-formatter | When TOON helps, when it does not, and how to invoke the TOON commands. |
Agent Targets
| Agent | Generated output | Notes |
|---|---|---|
| Claude Code | .claude/ |
Native Claude skills, settings, TOON slash commands, optional hooks. |
| Codex | AGENTS.md + .codex/skills/*/SKILL.md |
Root Codex guidance points to project-local skill files. |
| Cursor | .cursor/rules/*.mdc |
Cursor project rules generated as Agent Requested rules, plus an always-applied skill-selection rule. |
Claude remains the default for backwards compatibility. Use --agent all to install all supported targets.
Install
# Claude Code only (default)
npx [email protected]
# Codex only
npx [email protected] --agent codex
# Cursor only
npx [email protected] --agent cursor
# Claude Code + Codex + Cursor
npx [email protected] --agent all
# Backwards-compatible aliases still work
npx [email protected] --agent all
For Claude TOON commands, add the runtime deps to your project:
npm i @toon-format/toon gpt-tokenizer
Profiles
npx [email protected] --profile web-saas --agent all
npx [email protected] --profile fintech --agent codex,cursor
npx [email protected] --skills stripe,copywriting-frameworks --agent cursor
Profiles select a skill set. Agent targets decide where that skill set is installed.
Structure
.claude/
skills/<skill>/skill.md
commands/
utils/toon/cli.mjs
.codex/
skills/<skill>/SKILL.md
AGENTS.md
.cursor/
rules/agent-starter.mdc
rules/<skill>.mdc
rules/<skill>/references/
The package keeps one shared source of truth in templates/.claude/skills/ and generates Codex/Cursor formats from that source during install.
Documentation Pulls
The docs command is Claude-specific because external docs are stored under .claude/skills/*/docs.
npx [email protected] docs pull stripe
npx [email protected] docs status
Benchmarks
Real measured token counts for representative workloads are in bench/. Numbers use gpt-tokenizer, not a claimed heuristic.
Requirements
- Node.js >= 18
- Claude Code, Codex, or Cursor, depending on the selected target
- Optional:
@toon-format/toonandgpt-tokenizerfor Claude TOON slash commands
License
MIT. Not affiliated with Stripe, Supabase, Plaid, Expo, Anthropic, OpenAI, Cursor, or @toon-format/toon.
Yorumlar (0)
Yorum birakmak icin giris yap.
Yorum birakSonuc bulunamadi