Skip to content
Skribly Docs

Getting started

Each Skribly tool works on its own, and they work best together. This page takes you through the whole loop in about fifteen minutes. If you only want one tool, jump to its page:

  • Skribly Agent if you want a desktop app that runs your coding agents side by side.
  • agent-skills if you already use Claude Code, Codex, Cursor, Gemini CLI, OpenCode, Windsurf or GitHub Copilot and want them to follow one workflow.
  • Skribly Tracker if you want an issue tracker whose API is designed for agents.

What you need

Section titled “What you need”
What Why
git in your PATH Branches, worktrees, commits and downloading skills
At least one coding agent, for example npm install -g @anthropic-ai/claude-code Skribly runs the agents you already have, with their own logins
Node.js 20 or newer Only for agent-skills without Skribly Agent; the desktop app brings its own runtime
Docker Only to host Skribly Tracker yourself

The whole loop

Section titled “The whole loop”

  1. Start a tracker (optional). Run Skribly Tracker with Docker, as in Self-hosting, and create a personal access token on Profile → API tokens. You can skip this step and use Jira, Redmine, GitHub, GitLab or Linear instead.

  2. Install Skribly Agent. Download the file for your system and install it, as in Install Skribly Agent. The first launch opens a short tour; you can replay it from Help.

  3. Add a project. Open the project switcher in the top bar and choose Add project…. Skribly detects the base branch, the check commands and the dev URL of the repository.

  4. Connect a tracker. Go to Settings → Task trackers → Add tracker, choose the tracker, paste the URL and the token, then Save and check. Skribly Tracker is the default; locally its URL is http://localhost:8100.

  5. Install skills. Open Skills, download the agent-skills repository and install the skills you want for each agent in the project. Without the app, use the CLI as in Install agent-skills.

  6. Start a task. Open Tasks, pick a task and click Start with an agent. Skribly writes the task to .skribly/tasks/<KEY>/TASK.md, creates STATE.md and a branch from your pattern (or a worktree), and starts the agent with a ready prompt. The agent reads the task, proposes a plan and waits for your “ok”.

  7. Review and commit. Open Code & git, stage the files, click Suggest and then Commit. The message follows your repository’s history and never gets AI trailers.

  8. Close the task. Add a comment, change the status and log the time from the task panel, without leaving the app.

Without the desktop app

Section titled “Without the desktop app”

The same loop works in any agent’s own terminal with agent-skills:

Terminal window
git clone https://github.com/skribly/agent-skills.git ~/agent-skills
cd ~/agent-skills && npm link
agent-skills install --target all
agent-skills config attribution off
agent-skills doctor

Then, in a project, run /task 123 in Claude Code, $task 123 in Codex or /task 123 in the other agents. The task skill fetches the issue, writes .tasks/123/TASK.md, plans, branches after your approval, implements in stages and commits after your OK.

Next steps

Section titled “Next steps”