A terminal interface for the Skene ecosystem. Guides you through selecting a repository, choosing AI providers, and running Skene open-source libraries -- all from the terminal. Built with Go and Bubble Tea.
Skene CLI is the interactive front-end that installs and runs three Skene packages in your repository via uvx:
| Package | Purpose |
|---|---|
| Skene Growth | Tech stack detection, growth feature discovery, revenue leakage analysis, growth plan generation |
| Skene Skills | PLG analysis skills for Claude Code -- analyze, generate manifests and templates |
| Skene Cookbook | 700+ AI skills for PLG, marketing, security, DevEx, and more |
By default all three packages are used during analysis. You can choose a subset in the advanced configuration screen.
The CLI itself does not perform any analysis -- it orchestrates the external libraries by running them via uvx in your selected repository directory.
- Step-by-step setup -- provider, model, authentication, and project selection
- Multiple AI providers -- Skene, OpenAI, Anthropic, Gemini, Ollama, LM Studio, or any OpenAI-compatible endpoint
- Authentication flows -- Skene magic link, API key entry, and local model auto-detection
- Skene package selection -- choose which Skene packages to include (Growth, Skills, Cookbook)
- Existing analysis detection -- detects previous
skene-context/output and offers to view or rerun - Multi-phase analysis -- animated progress with live terminal output
- Tabbed results dashboard -- view growth plans, manifests, and product docs
- Next steps menu -- run
plan,build,validate, or re-analyze directly from the CLI - Delegates to libraries -- all analysis runs via
uvx skene-growthin your selected repo - Error handling -- severity-based error display with suggestions, retry, and go back
- Cross-platform -- macOS, Linux, and Windows
- Mini-game -- space shooter while you wait
None. No Go, Python, or other tools need to be installed. The Makefile downloads a pre-built binary if Go is not present, and the CLI automatically downloads the uv runtime on first use.
git clone https://github.com/Px8-fi/skene-cli.git
cd skene-cli
make build # builds from source if Go is installed, otherwise downloads pre-built binary
make runThis works on macOS (Intel + Apple Silicon), Linux (x86_64), and Windows (x86_64).
curl -fsSL https://raw.githubusercontent.com/Px8-fi/skene-cli/main/install.sh | bashgit clone https://github.com/Px8-fi/skene-cli.git
cd skene-cli
make install # download Go dependencies
make build # build from source
make runmake install-bin # copies build/skene to /usr/local/binRun skene and follow the prompts. The flow is:
Welcome
-> AI Provider selection (Skene, OpenAI, Anthropic, Gemini, Ollama, LM Studio, Generic)
-> Model selection
-> Authentication (magic link, API key, or local model detection)
-> Project directory (detects existing analysis if present)
-> Analysis configuration (default or custom package selection)
-> Running analysis (uvx skene-growth analyze . with live progress)
-> Results dashboard (Growth Plan | Manifest | Product Docs)
-> Next steps
When you select a project directory that already contains a skene-context/ folder, you are given two options:
- View Analysis -- loads the existing results into the dashboard
- Rerun Analysis -- proceeds to configuration and runs a fresh analysis
The default configuration runs all three Skene packages. Selecting "No" on the default question opens the advanced screen where you can toggle individual packages (Skene Growth, Skene Skills, Skene Cookbook).
After analysis completes, the next steps menu offers:
| Action | Description |
|---|---|
| Generate Growth Plan | Prioritised growth plan with implementation roadmap |
| Build Implementation Prompt | Ready-to-use prompt for Cursor, Claude, or other AI tools |
| Validate Manifest | Validate the growth manifest against the schema |
| Re-run Analysis | Analyse the codebase again |
| Open Generated Files | Open ./skene-context/ in your file manager |
| Change Configuration | Return to provider selection |
| Exit | Close Skene CLI |
All commands run via uvx skene-growth in your selected repository directory.
| Key | Action |
|---|---|
Up/Down or j/k |
Navigate |
Left/Right or h/l |
Navigate / switch tabs |
Enter |
Confirm / select |
Esc |
Go back |
Tab |
Switch focus area |
Space |
Toggle checkbox / option |
? |
Toggle help overlay |
g |
Play mini-game (during loading) |
Ctrl+C |
Quit |
Configuration files are checked in order (first found wins):
- Project --
.skene.configin the project directory - User --
~/.config/skene/config
Example .skene.config:
{
"provider": "gemini",
"model": "gemini-3-flash-preview",
"api_key": "your-api-key",
"base_url": "",
"output_dir": "./skene-context",
"verbose": true,
"use_growth": true,
"use_skills": true,
"use_cookbook": true
}| Provider | ID | Auth | Notes |
|---|---|---|---|
| Skene | skene |
Magic link or API key | Recommended -- built-in growth model |
| OpenAI | openai |
API key | GPT-4o, GPT-4 Turbo, GPT-3.5 Turbo |
| Anthropic | anthropic |
API key | Claude Sonnet 4.5, Opus, Haiku |
| Gemini | gemini |
API key | Gemini 3 Flash, 3 Pro, 2.5 Flash |
| Ollama | ollama |
None (local) | Llama 3.3, Mistral, CodeLlama, DeepSeek R1 |
| LM Studio | lmstudio |
None (local) | Uses currently loaded model |
| Generic | generic |
API key + base URL | Any OpenAI-compatible endpoint |
skene-cli/
├── cmd/skene/
│ └── main.go # Entry point
├── internal/
│ ├── tui/
│ │ ├── app.go # State machine and message handling
│ │ ├── styles/
│ │ │ └── styles.go # Colour palette and Lip Gloss styles
│ │ ├── components/
│ │ │ ├── button.go # Button and ButtonGroup
│ │ │ ├── dir_browser.go # Directory browser
│ │ │ ├── help.go # Help overlay and footer
│ │ │ ├── logo.go # ASCII logo
│ │ │ ├── ascii_motion_placeholder.go
│ │ │ ├── progress.go # Progress bar
│ │ │ ├── terminal_output.go # Scrollable terminal output
│ │ │ └── wizard_header.go # Step progress header
│ │ └── views/
│ │ ├── welcome.go # Welcome screen
│ │ ├── syscheck.go # System prerequisite checks
│ │ ├── install_method.go # uvx vs pip selection
│ │ ├── installing.go # Installation progress
│ │ ├── provider.go # AI provider selection
│ │ ├── model.go # Model selection
│ │ ├── auth.go # Skene magic link auth
│ │ ├── apikey.go # API key entry
│ │ ├── local_model.go # Local model detection
│ │ ├── project_dir.go # Project directory + existing analysis detection
│ │ ├── analysis_config.go # Package selection and options
│ │ ├── analyzing.go # Live analysis progress
│ │ ├── results.go # Tabbed results dashboard
│ │ ├── next_steps.go # Post-analysis actions
│ │ └── error.go # Error display
│ ├── services/
│ │ ├── auth/callback.go # OAuth callback server
│ │ ├── config/manager.go # Config file management
│ │ ├── growth/
│ │ │ └── engine.go # uvx command spawner
│ │ ├── ide/communicator.go # IDE integration
│ │ ├── syscheck/checker.go # System checks
│ │ └── uvresolver/resolver.go # Auto-provisions uv binary
│ └── game/
│ └── shooter.go # Space shooter mini-game
├── designs/ # Design reference images
├── Makefile
├── go.mod
└── go.sum
Skene CLI acts as an orchestrator. It does not contain analysis logic. Instead, it:
- Guides the user through provider, model, and project selection
- Auto-provisions
uvif not already installed (downloads to~/.skene/bin/on first use) - Spawns
uvx skene-growth analyze .(and other commands) in the selected repository directory - Streams output back to the TUI for live progress display
- Reads results from
{repo}/skene-context/and displays them in the dashboard
Provider, model, and API key are passed to uvx skene-growth via CLI flags and environment variables.
The application is a finite state machine driven by Bubble Tea messages:
Welcome -> Provider -> Model -> Auth/APIKey -> ProjectDir -> AnalysisConfig -> Analyzing -> Results -> NextSteps
| |
LocalModel (existing analysis?)
/ \
View Results Rerun -> AnalysisConfig
Every view in internal/tui/views/ implements a consistent interface:
SetSize(width, height int)
Render() string
GetHelpItems() []HelpItemThe main app.go orchestrates state transitions and delegates key handling to view-specific methods.
Lip Gloss with a warm, retro terminal palette:
- Cream/Amber
#EDC29C-- primary accent - Dark background
#1A1A1A - Success
#7CB374, Warning#E6B450, Error#F05D5E
Light terminal backgrounds are detected automatically and colours are adjusted for contrast.
Errors are categorised by severity (Warning, Error, Critical) and displayed in a consistent view with:
- Severity indicator and error code
- Clear message and suggested fix
- Retry / Go Back / Quit buttons
- Press
Escto go back to the previous screen
# Live reload (requires air)
make dev
# Run tests
make test
# Lint
make lint
# Format
make fmt
# Build for all platforms
make build-all
# Package releases
make release- Bubble Tea -- TUI framework
- Lip Gloss -- Styling
- Bubbles -- UI components (textinput, viewport)
- pkg/browser -- Browser opening for magic link auth
- termenv -- Terminal capability detection
- skene-growth -- PLG analysis toolkit (CLI + MCP server)
- skene-skills -- Claude Code plugin for PLG analysis
- skene-cookbook -- 700+ AI skills for Claude and Cursor
MIT