baseline-core

KO | EN

gitlite — search

#python #javascript #typescript #vue #python3 #css #library #docker #react #nodejs #machine-learning #golang

★ 24 Open GitHub ↗

Open-source skills system that wires your business context into AI tools

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# The Baseline System

[![npm version](https://img.shields.io/npm/v/@baseline-studio/cli.svg)](https://www.npmjs.com/package/@baseline-studio/cli)
[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
[![GitHub stars](https://img.shields.io/github/stars/TrentM6/baseline-core.svg?style=social)](https://github.com/TrentM6/baseline-core)

> A complete AI system for product teams. Skills provide methodology. Context makes it yours. Frameworks give structure.

Open source. MIT licensed. [baselinestudio.design](https://baselinestudio.design)

## Quick Start

**New users:**
```bash
npx @baseline-studio/cli init
```

`init` asks for your company name, scaffolds the system (skills, frameworks, context templates), runs `npm install`, and commits a fresh git repo. ~60 seconds.

**Then:** open the folder in your AI tool (Claude Code, Cursor, Codex, etc.) and say *"run the setup skill."* Setup walks you through filling in your business context — bring any docs you already have (pitch deck, one-pager, brand guide) and it'll parse them into your context files, asking only for what's missing.

**Existing users:**
```bash
npx baseline status    # Check for updates
npx baseline update    # Pull latest version
```

---

## What This Is

The Baseline System is an AI-powered workflow system that helps product teams do research, strategy, design, communication, and delivery — faster and at a higher quality than working without it.

It works with Claude Code (and other AI tools) by giving them structured knowledge about how to do product work and specific knowledge about your business.

**Three components:**

| Component | What It Does | Details |
|-----------|-------------|---------|
| **Skills** | Domain expertise modules that teach AI how to execute specific work | [`skills/_README.md`](skills/_README.md) |
| **Context** | Your business-specific knowledge — identity, voice, customers, product | `context/` |
| **Frameworks** | Reusable methodologies — prioritization, research, decision-making | [`frameworks/_README.md`](frameworks/_README.md) |

Skills are universal. Context is yours. When you run a skill, it loads both — producing output that follows proven methodology and sounds like your brand.

---

## How It Works

1. **You describe a task** — "Write a PRD for the new onboarding flow" or "Help me prioritize Q2 roadmap"
2. **The system identifies the skill** — PRDs map to `product-communications`, prioritization maps to `strategic-advisory`
3. **The skill loads its files** — The skill's `manifest.yaml` lists what to load: the skill file, frameworks, and your context
4. **The system executes** — Plan, clarify, execute, validate

In AI coding tools, this is fully automated via the `AGENTS.md` file. Just describe what you need.

### The Skills

**Run once to set up:**

| Skill | What It Does |
|-------|-------------|
| **Setup** | Guided context setup — ingest your existing docs, parse them into context files, ask gap-filling questions for what's missing |

**Run as needed for daily work:**

| Skill | What It Does |
|-------|-------------|
| **Strategic Advisory** | Strategic decisions, roadmaps, prioritization, OKRs, alignment |
| **Research & Synthesis** | User research, interviews, competitive analysis, synthesis |
| **Product Communications** | PRDs, feature specs, product briefs, stakeholder updates, decision docs |
| **UX Design** | Interface design, wireframes, user flows, UI copy, design systems, accessibility |
| **Product Analytics** | Metrics, dashboards, funnel analysis, A/B tests, segmentation |
| **Prototyping** | Coded prototypes, clickable demos, POCs, technical feasibility |
| **Project Management** | Planning, tracking, sprints, status updates, risk management |
| **Technical Documentation** | User guides, help center, API docs, release notes, how-to guides |
| **Visual Communication** | Presentations, diagrams, decision visualization, data storytelling, narrative decks |
| **Product Marketing** | Positioning, messaging frameworks, launch briefs, competitive messaging, case studies |
| **Go-to-Market Planning** | Pricing strategy, launch planning, channel strategy, competitive positioning |
| **Skill Building** | Create new skills, document expertise, build reference files |

See [`skills/_README.md`](skills/_README.md) for detailed documentation on each skill.

---

## Using the System

### AI Coding Tools (Automated)

`AGENTS.md` is the canonical instruction file. It tells AI tools how to find skills, load manifests, read context, and execute workflows. Here's how each tool picks it up:

| Tool | How It Works |
|------|-------------|
| **Codex, Cursor, Windsurf, JetBrains AI** | Reads `AGENTS.md` directly |
| **Claude Code** | Reads `CLAUDE.md`, which points to `AGENTS.md` |
| **GitHub Copilot Chat** | Reads `.github/copilot-instructions.md`, which points to `AGENTS.md` |

1. Open your system folder in your AI coding tool
2. Describe what you need
3. Done — the tool reads `AGENTS.md` (directly or via pointer) and handles the rest

### ChatGPT / Gemini / Other Chat Tools

Chat tools can't auto-read project files. Upload or paste the files listed in the skill's `manifest.yaml` into your conversation, then describe your task.

---

## Context

Context is what makes the system yours. The system's output quality is directly proportional to the quality of your context.

**The fastest way to fill it in:** run the **Setup** skill in your AI tool. It asks for any docs you have (pitch deck, brand guide, etc.), parses them into your context files, then asks short follow-ups for whatever's missing. Identity + voice + co-founder persona — enough to start using the system meaningfully — typically takes ~10 minutes.

You may know this pattern as soul.md — a single file that defines who your AI is. Baseline's context layer is the same idea, broken into modules. Identity, voice, personas, competitive positioning, and business knowledge each get their own file, so different skills can load exactly what they need. Same principle. More control.

**Core context** (loaded by every skill):
- `context/core/identity.md` — Who you are, what you do, positioning, differentiators
- `context/core/voice.md` — How you sound, language rules, tone by channel

**Extended context** (loaded by relevant skills):
- `context/extended/product.md` — Product details, features, workflow
- `context/extended/users.md` — User personas, goals, pain points
- `context/extended/icp.md` — Ideal customer profile, buyer psychology
- `context/extended/competitive.md` — Competitors, positioning, alternatives
- `context/extended/pricing.md` — Pricing tiers, objection handling
- `context/extended/technical.md` — Tech stack, integrations, constraints
- `context/extended/visual-identity.md` — Colors, fonts, visual style
- `context/extended/formatting.md` — Document structure, heading rules
- `context/extended/proof-points.md` — Metrics, case studies, market research, validation data

Skills work with whatever context is available. Missing context produces more generic output, but nothing breaks — fill what matters most first.

**Prefer the terminal?** Run `npx baseline context` for a guided CLI questionnaire (alternative to the Setup skill). Use `npx baseline context add <name>` to create a new context file.

---

## CLI Reference

See [`cli/README.md`](cli/README.md) for the full CLI documentation.

| Command | What It Does |
|---------|-------------|
| `npx @baseline-studio/cli init` | Create a new Baseline System |
| `npx baseline status` | Show current version, check for updates |
| `npx baseline update` | Pull latest skills, frameworks, and CLI |
| `npx baseline context` | Terminal questionnaire for filling context (alternative to the Setup skill) |
| `npx baseline context add <name>` | Create a new context file and wire it to skills |

---

## File Structure

```
your-system/
├── AGENTS.md                    # AI instructions for all coding tools (don't edit)
├── CLAUDE.md                    # Claude Code pointer to AGENTS.md
├── .github/
│   └── copilot-instructions.md  # GitHub Copilot pointer to AGENTS.md
├── baseline.config.json         # Version tracking and config
├── skills/                      # 13 domain expertise modules (Setup + 12 daily-work skills)
├── context/                     # YOUR business knowledge (you own this)
│   ├── core/                    # Identity and voice (loaded by every skill)
│   └── extended/                # Product, users, pricing, etc.
├── frameworks/                  # Reusable methodologies
└── cli/                         # Bundled CLI for daily use
```

**You own `context/`.** Everything else is managed by the system and updated via `npx baseline update`.

---

## FAQ

**Can I edit skill files?**
You can, but changes are overwritten on `baseline update`. Put custom behavior in context files instead.

**What if I don't have all the context files filled out?**
Skills work with whatever context is available. Missing context means more generic output, but nothing breaks.

**Can I add my own context files?**
Yes. Run `npx baseline context add <name>` to create a new file and wire it to skills.

**What AI tools does this work with?**
Any AI tool. The `AGENTS.md` file is automatically read by Claude Code, Codex, Cursor, Windsurf, GitHub Copilot, JetBrains AI, and others. Chat tools like ChatGPT work when you upload skill and context files.

**How do I contribute?**
See [CONTRIBUTING.md](CONTRIBUTING.md). Open an issue first, then submit a PR.

**How do I get help?**
[Open an issue](https://github.com/TrentM6/baseline-core/issues) or email trent@baselinestudio.design.

---

Baseline Studio researches how product teams can work better with AI. Built by [Trent](https://www.linkedin.com/in/trentjmitchell/).

baseline-core

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README