| name | using-xskill |
| description | Use when installing, configuring, or operating xskill (the `xskill` CLI / `pip install xskill`) — starting the daemon, registering trajectory dirs, joining a team server, understanding how trajectories become Skills, or rebuilding/re-distilling the skill library after a model change. |
Using xskill
Overview
xskill distills reusable Skills (SKILL.md folders) out of the real execution
trajectories of coding agents (Claude Code, Codex, OpenCode, Cursor, …). A background
daemon watches each agent's session logs, slices them into single-intent atoms,
clusters atoms into skills, and writes/versions each skill in its own git folder. New
skill versions only replace old ones when real traffic shows they serve users better
(canary A/B by UX score) — not by an LLM grading itself.
Core mental model: raw trajectory → atoms → candidate routing → SKILL.md → canary
A/B → installed into every agent's skill dir. You operate the daemon; the daemon does
the distilling.
When to Use
- Installing xskill or filling in
~/.xskill/config.yaml (LLM + embedding endpoints)
- Starting/keeping the daemon running (
xskill serve), or backfilling old trajectories
- Joining or hosting a team server (
xskill serve --server / xskill connect)
- Understanding the agent pipeline, atoms, canary/UX scoring, or deployment modes
- Re-distilling the whole skill library (e.g. after switching to a stronger model)
Quick Reference
| Command | What it does |
|---|
pip install xskill | Install (Python 3.9+) |
xskill serve | Standalone daemon: FastAPI + watcher; first run writes ~/.xskill/config.yaml then exits |
xskill serve --server | Team server: owns all LLM calls + git; prints a join token |
xskill connect <host:port> --token <t> | Join a team server as a thin client |
xskill registry add <path> | Backfill / watch an extra trajectory directory |
xskill search traj|skill <query> | Search trajectories or skills |
xskill read <path> --eco <eco> | Batch-ingest db trajectories (ngagent/opencode) |
xskill rebuild [--force] | Re-distill from existing raw trajectories (see reference) |
xskill stats | Token usage & estimated cost |
The daemon is the engine: most commands only change state in the DB; nothing is
distilled unless xskill serve (or the team server) is running.
Progressive Disclosure — read on demand
- Install & configure (config.yaml fields, per-agent collect/install paths, team
client setup):
references/installation.md
- How it works (TaskAgent → TaskClusterAgent → SkillEditAgent, atoms, canary/UX
scoring, standalone vs team mode):
references/mechanisms.md
- Rebuild the skill library (a ready-to-run prompt that walks a model through
re-distilling correctly):
references/rebuilding-skill-library.md
Common Mistakes
- Running
rebuild with no daemon up. rebuild only resets DB state; the watcher
in serve does the actual re-split/re-cluster every 30s. No daemon = nothing happens.
- Deleting raw
~/.xskill/*_sessions/*.md. Those are the input to distillation —
delete them and you can no longer rebuild.
- Expecting DeepSeek to do embeddings. DeepSeek has no embedding endpoint; point the
embedding: block at DashScope / OpenAI / Ollama.
- Putting tokens in public places. Team join tokens must never land in a public repo
or chat log.
