xskill CLI

The xskill CLI gathers sanitized session trajectories from Claude Code, Codex and OpenCode and grows a shared skill library.

  • 129 ★ on the repository
  • Last commit
  • MIT
CLI

Extended access

Why this level

  • The tool reads session files of every agent on the machine and in team mode uploads sanitized trajectories to a server.
  • Automatically generated skills land in agent directories and change their behavior, so versions need review.
All reasons and checks

skillnerds/xskill

Install

In your terminal, with SkillFoxx CLI

npx skillfoxx add cli/xskill

Detects the agents on your machine, shows the access level and pins the version.

Other ways to install

Access and checks

How we review

Extended accessThe tool runs code or makes network calls and only reads external data.

  • The tool reads session files of every agent on the machine and in team mode uploads sanitized trajectories to a server.
  • Automatically generated skills land in agent directories and change their behavior, so versions need review.

This is third-party code. Review the repository files before installing.

What it does

xskill watches agent sessions on a machine, splits each into sub-tasks, clusters them and, when enough accumulate, writes a new skill version automatically. Finished skills return to the agents' directories: Claude Code, Codex, OpenCode and others. If one person solved a task in their session, the rest get that skill. There are two modes. In single mode xskill searches your past sessions with `xskill traj search`. In team mode a server collects colleagues' trajectories, selects skills by profile and keeps versions scored on real runs. For distillation an LLM and an embedding model are set in `~/.xskill/config.yaml`.

Who it is for. Development teams whose colleagues use Claude Code, Codex or Cursor and keep repeating the same solutions.

Good fit when

  • An experienced colleague solved a tangled migration and the others need that path as a skill.
  • A memory leak was already investigated and the old session has to be found by keywords.
  • Team skills go stale and a score from real runs with a version chart is needed.
  • Skills from an internal hub must be included in recommendations and scoring.

Not a fit when

  • Agent conversations must not be uploaded to a shared server even after sanitizing.
  • There are no keys for an LLM and an embedding model: without them skills are neither grouped nor written.
  • The team is one person with no repeating workflows: search over your own sessions is enough.

Example request

Find the Python memory leak investigation in session history and show the steps that worked.

Limitations

Python 3.9 or newer is required. For skill distillation `config.yaml` needs the addresses and keys of two models, an LLM and an embedding one; without them only session search works. Single mode loses many team features, so it is recommended only for people who want to turn frequent workflows into skills or run agents for long every day. By default trajectories of all projects go to the server; narrow the upload with `xskill privacy mode allowlist`. Claude Code, Codex and OpenCode are verified, while OpenClaw is implemented but not fully verified.

How to disable. Stop `xskill serve` and remove the package with `pip uninstall xskill`. Delete the skills the tool symlinked into agent directories and the `~/.xskill/` folder separately.

FAQ

Where do I start on one machine?

Run `pip install xskill`, then `xskill init`: the tool finds agents and converts sessions. Search with `xskill traj search <word>` and read with `xskill traj read <traj_id>`.

How do I join a team server?

The administrator runs `xskill serve --server` and receives a join command for colleagues. A member runs `xskill connect <host:port> --token <token> --name <name>`.

How do I restrict trajectory upload?

Enable the allowlist with `xskill privacy mode allowlist`, then run `xskill privacy allow` in each project folder. `xskill privacy status` shows each project.

Can I generate a skill manually?

Yes. In team mode `xskill generate "<description>"` creates or rewrites a skill from collected trajectories, and `--name alice,bob` limits it to trajectories of specific people.

How do I download a ready skill?

Find it with `xskill search <word>`, then `xskill download <skill-id> --agent claude-code -y` installs it into the chosen agent.

Discussion

No questions or reviews yet.

CLI tools

A desktop app and CLI to install, sync and organize agent skills across Claude Code, Codex, Cursor and other tools

CLIExtended access7.5KRepository stars
Official

A skill registry for teams with security scanning and versioning; the MCP server finds and installs skills on request.

MCP serverExtended access78Repository stars

A meta-skill that takes a vague idea to a finished agent skill: it designs, writes the files, checks them and refines by evidence

SkillExtended access250Repository stars

The skillx CLI searches a skills catalog, vets skills before install, pins versions in a lock file and serves an MCP installer.

CLIExtended access197Repository stars

A CLI and manager skill: one skills library distributed to 70+ agents, with management done from the agent conversation

CLIExtended access243Repository stars

A skill for creating skills that proves them by behavior: before-and-after runs, regression evals and ecosystem audit

SkillExtended access918Repository stars

README in short

The authors report gains over SkillOpt on three benchmarks: Spreadsheet, ALFWorld and OfficeQA. Skill versions are tested on live traffic, the best becomes primary and any change can be rolled back.

README badge

Report a problem