# thinking-in-systems

AI-assisted development framework for [Claude Code](https://claude.ai/code) and [Cursor](https://cursor.sh). Ships 29 skills that guide AI through your entire product lifecycle — from idea validation to growth.

## Quick Start

```bash
# Create a new project (no install needed)
npx thinking-in-systems init

# Or install globally for the short command
npm i -g thinking-in-systems
tis init my-app --type web
```

## What You Get

When you run `tis init`, your project gets:

```
my-app/
├── CLAUDE.md          # AI context file (auto-loaded by Claude Code)
├── .cursorrules       # Cursor config
├── .mcp.json          # MCP server config (11 servers)
├── skills/            # 29 AI development skills
│   ├── verification/          # Anti-hallucination protocol
│   ├── test-driven-development/  # Strict TDD enforcement
│   ├── systematic-debugging/     # 4-phase root cause analysis
│   ├── code-review/              # Two-stage review + anti-sycophancy
│   ├── phase-0-validation/       # Market research + kill checks
│   ├── phase-6-launch/           # Landing page, SEO, distribution
│   ├── phase-7-growth/           # Content, hooks, analytics
│   └── ...14 more
├── docs/              # PRD, tasks, changelog templates
├── brain-dump.md      # Start here — dump your idea
└── .github/           # CI/CD workflow templates
```

Then open Claude Code and say: *"Read brain-dump.md and run Phase 0 validation"*

## Commands

### `tis init [name]`

Create a new project with all skills and templates.

```bash
tis init                          # interactive
tis init my-app --type web        # non-interactive
tis init my-app --type mobile     # React Native/Expo
tis init my-app --type macos      # SwiftUI
tis init my-app --type api        # Backend service
tis init my-app --type fullstack  # Full-stack
```

### `tis add <skill...>`

Add specific skills to an existing project.

```bash
tis add tdd debugging             # add by name or alias
tis add verification code-review  # multiple at once
tis add --all                     # add everything
```

Dependencies are auto-resolved — adding a Phase 7 skill automatically copies the psychology frameworks it references.

### `tis list`

See all available skills.

```bash
tis list                       # all 29 skills
tis list --category core       # just core (14 skills)
tis list --category phase-0    # validation skills
tis list --category phase-7    # growth skills
```

### `tis mcp`

Setup the 8 global MCP servers that power the skills.

```bash
tis mcp          # install servers (prompts for API keys)
tis mcp --list   # check what's installed
```

**Free (no key needed):** Context7, Playwright, Dialog MCP, Google News Trends
**Need API key (free tier):** Brave Search, Exa, Firecrawl, Apify

## The 29 Skills

### Core Development (14)

| Skill | What It Does |
|-------|-------------|
| `verification` | Anti-hallucination protocol with Iron Law gate — bans "should work" claims |
| `test-driven-development` | Strict RED-GREEN-REFACTOR. Wrote code before test? Delete it. |
| `systematic-debugging` | 4-phase root cause analysis. 3+ fixes fail = question the architecture |
| `code-review` | Two-stage review (spec compliance + code quality) with anti-sycophancy |
| `testing` | Unit, integration, E2E with property-based testing triggers |
| `context-management` | Session state handoffs when context window fills up |
| `architecture-design` | Tech stack decisions |
| `frontend-design` | UI/UX with Vercel design guidelines |
| `product-development` | Full 8-phase workflow |
| `git-workflow` | Branching, commits, PRs |
| `deployment` | Dev to production pipeline |
| `research` | MCP-powered market/tech validation |
| `launch-prep` | Pre-launch security + readiness checklist |
| `skill-creator` | Create your own skills (with TDD methodology) |

### Phase 0: Validation (4)

Validate your idea before building. Uses MCP servers for real data.

`brain-dump` → `market-research` → `competitor-analysis` → `kill-check` (GO/PIVOT/KILL)

### Phase 6: Launch (4)

Ship and distribute.

`landing-page` → `seo-aso` → `distribution` → `launch-kill-check`

### Phase 7: Growth (7)

Data-driven growth with psychology frameworks and slop detection.

`data-mining` → `content-research` → `hook-generation` → `content-creation` → `seo-optimization` → `performance-analysis` → `growth-kill-check`

### Frameworks

- `_psychology/` — Hook frameworks (PAS, Curiosity Gap, FOMO), emotional triggers, platform algorithms
- `_engine/` — Kill-check base scoring, token budgeting, safety rules

## How Skills Work

Skills are markdown files that Claude Code and Cursor read automatically. They contain:

- **When to use** — trigger conditions
- **Process** — step-by-step workflow
- **Rationalizations to reject** — agent loophole prevention
- **Pre-completion checklist** — gate before claiming done

You reference them in conversation: *"Use the verification skill to check this code"* or they activate automatically when Claude reads your `skills/` directory.

## Skill Aliases

Short names for faster `tis add`:

| Alias | Skill |
|-------|-------|
| `tdd` | test-driven-development |
| `debug` | systematic-debugging |
| `review` | code-review |
| `verify` | verification |
| `context` | context-management |
| `hooks` | hook-generation |
| `content` | content-creation |
| `seo` | seo-aso |

## Requirements

- Node.js >= 16
- Claude Code or Cursor (for using the skills)
- MCP servers (optional, for Phase 0/6/7 skills)

## License

MIT

---

Built by [@adeidei](https://github.com/adeidei)