Developer reference
Commands and permissions.
Choose the Skene surface you use to map the events Growth reports on or check a change against them. Commands copy as written, and the live database command is pinned to a verified source revision.
Run Skene OSS.
This Skene CLI reference covers three ways to run Skene OSS. Each one reads the inputs you choose and writes skene-context/journey.yaml with the model provider you configure.
Use uvx
With uv available, fetch and run the CLI without adding it to your application dependencies.
uvx skene analyse-journey .Install and open the TUI
The installer downloads the terminal application. The wizard guides provider configuration and analysis.
curl -fsSL https://raw.githubusercontent.com/SkeneTechnologies/skene/main/tui/install.sh | bash
skeneInstall the Python package
Install the CLI into your chosen Python environment, then run the analysis.
pip install skene
skene analyse-journey .Choose one schema input.
Exported SQL and a live database URL are alternatives. Use one flag, not both. Leave both out to run without a database schema input.
uvx skene analyse-journey . --schema-dir ./schemasThe schema agent reads the SQL files in the directory you supply.
uvx --from git+https://github.com/SkeneTechnologies/skene.git@8eb7942f569a4f99387e5c05b086c7babc8ce3d3 skene analyse-journey . --db-url "postgresql://USER:PASSWORD@HOST:5432/DATABASE"The published 0.5.1 package does not include the live database option. This command uses a verified source revision. Replace the placeholders locally; keep real database passwords out of shared code and screenshots.
Inspect the verified sourceUse either the live database option or exported SQL, not both.
Default output: skene-context/journey.yaml. Use --output to choose another path. See the full CLI reference for provider and configuration options.
GitHub App access.
Install on all repositories or a selected set, then link the workspace repository. The app never pushes directly to your default branch.
| Permission | Access | Purpose |
|---|---|---|
| Contents | Read & write | Read source and context files; propose updates through pull requests. |
| Pull requests | Read & write | Read changes, post reviews and open fix pull requests. |
| Issues | Read & write | Respond to /skene fix comments. |
| Metadata | Read | List repositories accessible to the installation. |
Hosted MCP access.
Use the hosted endpoint with a valid workspace credential or the documented authorization flow. Your agent sees only the tools its credential’s scopes allow.
https://www.skene.ai/api/mcpskene_check returns scan steps. skene_gap compares the supplied scan with the journey baseline. skene_analytics_review returns a model review of a diff without posting to GitHub.
The deterministic comparison and the GitHub review are different paths. A finding in code does not prove whether an event arrived in an analytics source.
See the check in a pull request.
Follow a removed tracking call from the code change to the engineer’s review.
Explore the developer workflow