Metadata-Version: 2.4
Name: corsette
Version: 0.2.3
Summary: Portable AI context product with OKF knowledge and generated adapters.
Requires-Python: >=3.11
Description-Content-Type: text/markdown
Requires-Dist: PyYAML>=6
Provides-Extra: dev
Requires-Dist: pytest>=8; extra == "dev"

# Corsette

> **Corsette turns one governed catalog of agents, skills, knowledge, and commands into a small, reproducible runtime for your AI tools.**

## What you get

- **80 agents**, **34 skills**, **34 commands**, and **17 MCP definitions**.
- One canonical catalog rendered for OpenCode, Claude Code, pi.dev, aicontext, and AGENTS.md runtimes.
- Global or project-local installation; generated output is disposable and private context stays local.
- Offline validation, secret scrubbing, redacted transcript support, and optional RTK shell-output compression.

## Install

```bash
git clone https://github.com/arturonaredo/corsette.git
cd corsette
python3 -m pip install -e '.[dev]'

# Generate and install the global OpenCode runtime (creates a timestamped backup first)
corsette generate --adapter opencode
corsette install --adapter opencode --global-runtime
```

For one repository instead:

```bash
corsette install --adapter opencode --local
```

## Use and evolve

Edit `catalog/` for agents, skills, commands, profiles, and MCPs; edit `okf/knowledge/` only for reusable public knowledge. Then validate and regenerate:

```bash
corsette validate
python tools/check-skills.py
corsette generate --adapter opencode
```

See [`catalog/agents`](catalog/agents), [`catalog/skills`](catalog/skills), and [`catalog/mcp`](catalog/mcp) for the complete catalog. Keep credentials in your local `.env`, never in Corsette.
