# Claude Code Agents, Skills & Slash Commands for SaaS Founders

> Install 46 Claude Code subagents, 61 skills, 47 slash commands, and 12 rules into .claude/. Generated from the same kit that ships for Cursor, so the two never drift.

Canonical: https://www.agentickit.co/claude-code
Last updated: 2026-11-18

46 subagents, 61 skills, 47 slash commands and 12 rules, installed into `.claude/`. Generated from the same source kit that ships for Cursor, so the two editions cannot drift apart.

## What lands in .claude/
- **46 subagents** - `.claude/agents/<slug>.md`, each with a `tools` and `model` grant scoped to what that role is allowed to do
- **47 slash commands** - `.claude/commands/<slug>.md`, with `description`, `argument-hint` and `allowed-tools`
- **61 skills** - `.claude/skills/<slug>/SKILL.md`, the playbooks agents must read before acting
- **12 rules** - `.claude/rules/<name>.md`, imported into every session by `CLAUDE.md`
- **Hooks** - `.claude/settings.json`, which keeps `CLAUDE.md` in sync with your stack on session start and after edits

## Install
```
npx agentickit-cursor login YOUR_LICENSE_KEY
npx agentickit-cursor init --target claude
```
`--kit engineering | marketing | both` selects *which* assets you get and is tied to your licence. `--target cursor | claude | both` selects *where* they are written and is free. The two compose.

## How the Claude Code kit is built
The `.cursor/` tree is the single source of truth. The Claude Code edition is generated from it by `scripts/build-claude-kit.js`, which rewrites paths, converts `.mdc` rules to `.md`, and attaches the Claude-native frontmatter each asset type needs. `--check` regenerates into a temporary directory and deep-diffs against what is committed; if a single byte differs, the build fails.

## What Claude Code gets that Cursor does not
Cursor agent files carry a name and a description. Claude Code subagents also carry a tool grant and a model, which turns a written instruction into an enforced boundary. Slash commands gain `argument-hint` and `allowed-tools` frontmatter. Rules ship as plain `.md` and are pulled into context by `CLAUDE.md` via `@.claude/rules/<name>.md` imports. Hooks live in `.claude/settings.json` instead of `.cursor/hooks.json`.

## What is identical
All 61 skills are the same files. The agent playbooks, the command pipelines, the stop conditions and the definitions of done are the same content across both editors.

## FAQ
### What lands in .claude/ when I install?
46 subagents in `.claude/agents/`, 47 slash commands in `.claude/commands/`, 61 skills in `.claude/skills/<slug>/SKILL.md`, and 12 rules in `.claude/rules/` imported by CLAUDE.md. Hooks are wired in `.claude/settings.json`.

### Is this the same kit as the Cursor one?
Yes. One kit emitted for two targets. Only the file format and frontmatter differ.

### Do the two versions drift apart over time?
No. `.cursor/` is the source of truth; `.claude/` is generated from it, and `--check` fails the build on any byte difference.

### Can I install Claude Code only?
Yes: `npx agentickit-cursor init --target claude` writes nothing to `.cursor/`.

### Will it overwrite a CLAUDE.md I already have?
No, if it has real content. It's only written when missing or still the placeholder.

## Related
- [All agents](https://www.agentickit.co/agents)
- [Install docs](https://www.agentickit.co/docs)
