Slancha
Slancha home

Install one skill. Know which version you have.

Give your coding agent a recipe it can follow. You choose the workspace and the skill, handle sign-in, and approve changes before they happen.

Claude Code · Skills setup recipe v1 · Verified 4 October 2026

Let your agent help with setup.

Copy the recipe into Claude Code. Tell it the exact name of a skill already approved in your workspace. It will inspect your setup, preview changes and stop for your approval.

Open text recipeDownload recipe

Prefer to set it up yourself?

  1. Check the prerequisites.

    Use your own terminal and user account. You need curl, Python 3.9+, a SHA-256 tool, Git and Claude Code. Choose an approved skill in your workspace before installing it.

  2. Read the installer, then run it.

    curl -fsSL https://app.slancha.ai/install.sh -o /tmp/slancha-install.sh
# Read the downloaded script before running it.
sh /tmp/slancha-install.sh

    The script verifies its pinned CLI download, installs under your user account and starts sign-in and setup. It installs the skill plugin without configuring tracing. If Slancha is already installed, use slancha init --dry-run to preview setup first.

  3. Finish sign-in yourself.

    Open the approval link, match the device code and choose your workspace. Never paste tokens into chat. Use the host-specific command in your console’s Claude Code setup page if it differs from the public installer.

  4. Preview a named skill, then install it.

    In your project directory, replace the value below with the exact published name from your library. Review the preview before running the second command.

    approved_skill='your-approved-skill-name'
slancha update "$approved_skill" --dry-run
slancha update "$approved_skill"
slancha status

    A pending skill must clear its scan and be approved by a person first. Keep local edits; do not force updates. Compare the installed revision with the published version. Reload skills or restart Claude Code when prompted.

Tracing is a separate choice.

Skills setup works without tracing. Enabling tracing needs a separate opt-in and administrator approval. Premium session mining is a separate capability and can send session excerpts to your workspace’s model provider after an admin acknowledges the notice.

Custom decision models are upcoming. This recipe installs skills; decision-model access and delivery will have their own setup when available.

Something did not complete?

Skill not published
Open Review in the console. An authorized person must review the scan and diff before approval.
Sign-in expired
Run slancha auth again and finish the new device-code approval yourself.
Missing or unknown revision
Do not mark setup complete. Check the named skill, local edits and connectivity with slancha status.
Recipe unavailable
Use the manual steps or the CLI documentation. No tracing or forced overwrite is needed.
Ask for help