using-xskill

Name: using-xskill
Author: SkillNerds

91stars12forks

How to install using-xskill?

skillnerds/xskill/using-xskill

$npx -y skills add SkillNerds/xskill --skill using-xskill

Installs into the current project.

Summary

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 cha

Files · 1

View on GitHub

SKILL.md

1	# Using xskill
2
3	## Overview
4
5	xskill distills reusable Skills (`SKILL.md` folders) out of the real execution
6	trajectories of coding agents (Claude Code, Codex, OpenCode, Cursor, …). A background
7	daemon watches each agent's session logs, slices them into single-intent atoms,
8	clusters atoms into skills, and writes/versions each skill in its own git folder. New
9	skill versions only replace old ones when real traffic shows they serve users better
10	(canary A/B by UX score) — not by an LLM grading itself.
11
12	Core mental model: raw trajectory → atoms → candidate routing → SKILL.md → canary
13	A/B → installed into every agent's skill dir. You operate the daemon; the daemon does
14	the distilling.
15
16	## When to Use
17
18	- Installing xskill or filling in `~/.xskill/config.yaml` (LLM + embedding endpoints)
19	- Starting/keeping the daemon running (`xskill serve`), or backfilling old trajectories
20	- Joining or hosting a team server (`xskill serve --server` / `xskill connect`)
21	- Understanding the agent pipeline, atoms, canary/UX scoring, or deployment modes
22	- Re-distilling the whole skill library (e.g. after switching to a stronger model)
23
24	## Quick Reference
25
26	\| Command \| What it does \|
27	\|---------\|--------------\|
28	\| `pip install xskill` \| Install (Python 3.9+) \|
29	\| `xskill serve` \| Standalone daemon: FastAPI + watcher; first run writes `~/.xskill/config.yaml` then exits \|
30	\| `xskill serve --server` \| Team server: owns all LLM calls + git; prints a join token \|
31	\| `xskill connect <host:port> --token <t>` \| Join a team server as a thin client \|
32	\| `xskill registry add <path>` \| Backfill / watch an extra trajectory directory \|
33	\| `xskill search traj\\|skill <query>` \| Search trajectories or skills \|
34	\| `xskill read <path> --eco <eco>` \| Batch-ingest db trajectories (ngagent/opencode) \|
35	\| `xskill rebuild [--force]` \| Re-distill from existing raw trajectories (see reference) \|
36	\| `xskill stats` \| Token usage & estimated cost \|
37
38	The daemon is the engine: most commands only change state in the DB; nothing is
39	distilled unless `xskill serve` (or the team server) is running.
40
41	## Progressive Disclosure — read on demand
42
43	- Install & configure (config.yaml fields, per-agent collect/install paths, team
44	client setup): `references/installation.md`
45	- How it works (TaskAgent → TaskClusterAgent → SkillEditAgent, atoms, canary/UX
46	scoring, standalone vs team mode): `references/mechanisms.md`
47	- Rebuild the skill library (a ready-to-run prompt that walks a model through
48	re-distilling correctly): `references/rebuilding-skill-library.md`
49
50	## Common Mistakes
51
52	- Running `rebuild` with no daemon up. `rebuild` only resets DB state; the watcher
53	in `serve` does the actual re-split/re-cluster every 30s. No daemon = nothing happens.
54	- *Deleting raw `~/.xskill/_sessions/.md`.* Those are the input to distillation —
55	delete them and you can no longer rebuild.
56	- Expecting DeepSeek to do embeddings. DeepSeek has no embedding endpoint; point the
57	`embedding:` block at DashScope / OpenAI / Ollama.
58	- Putting tokens in public places. Team join tokens must never land in a public repo
59	or chat log.

People who liked this also viewed

6 picks

Type

using-xskill

91stars12forks

View on GitHub

How to install using-xskill?

skillnerds/xskill/using-xskill

$npx -y skills add SkillNerds/xskill --skill using-xskill

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Using xskill
2
3	## Overview
4
5	xskill distills reusable Skills (`SKILL.md` folders) out of the real execution
6	trajectories of coding agents (Claude Code, Codex, OpenCode, Cursor, …). A background
7	daemon watches each agent's session logs, slices them into single-intent atoms,
8	clusters atoms into skills, and writes/versions each skill in its own git folder. New
9	skill versions only replace old ones when real traffic shows they serve users better
10	(canary A/B by UX score) — not by an LLM grading itself.
11
12	Core mental model: raw trajectory → atoms → candidate routing → SKILL.md → canary
13	A/B → installed into every agent's skill dir. You operate the daemon; the daemon does
14	the distilling.
15
16	## When to Use
17
18	- Installing xskill or filling in `~/.xskill/config.yaml` (LLM + embedding endpoints)
19	- Starting/keeping the daemon running (`xskill serve`), or backfilling old trajectories
20	- Joining or hosting a team server (`xskill serve --server` / `xskill connect`)
21	- Understanding the agent pipeline, atoms, canary/UX scoring, or deployment modes
22	- Re-distilling the whole skill library (e.g. after switching to a stronger model)
23
24	## Quick Reference
25
26	\| Command \| What it does \|
27	\|---------\|--------------\|
28	\| `pip install xskill` \| Install (Python 3.9+) \|
29	\| `xskill serve` \| Standalone daemon: FastAPI + watcher; first run writes `~/.xskill/config.yaml` then exits \|
30	\| `xskill serve --server` \| Team server: owns all LLM calls + git; prints a join token \|
31	\| `xskill connect <host:port> --token <t>` \| Join a team server as a thin client \|
32	\| `xskill registry add <path>` \| Backfill / watch an extra trajectory directory \|
33	\| `xskill search traj\\|skill <query>` \| Search trajectories or skills \|
34	\| `xskill read <path> --eco <eco>` \| Batch-ingest db trajectories (ngagent/opencode) \|
35	\| `xskill rebuild [--force]` \| Re-distill from existing raw trajectories (see reference) \|
36	\| `xskill stats` \| Token usage & estimated cost \|
37
38	The daemon is the engine: most commands only change state in the DB; nothing is
39	distilled unless `xskill serve` (or the team server) is running.
40
41	## Progressive Disclosure — read on demand
42
43	- Install & configure (config.yaml fields, per-agent collect/install paths, team
44	client setup): `references/installation.md`
45	- How it works (TaskAgent → TaskClusterAgent → SkillEditAgent, atoms, canary/UX
46	scoring, standalone vs team mode): `references/mechanisms.md`
47	- Rebuild the skill library (a ready-to-run prompt that walks a model through
48	re-distilling correctly): `references/rebuilding-skill-library.md`
49
50	## Common Mistakes
51
52	- Running `rebuild` with no daemon up. `rebuild` only resets DB state; the watcher
53	in `serve` does the actual re-split/re-cluster every 30s. No daemon = nothing happens.
54	- *Deleting raw `~/.xskill/_sessions/.md`.* Those are the input to distillation —
55	delete them and you can no longer rebuild.
56	- Expecting DeepSeek to do embeddings. DeepSeek has no embedding endpoint; point the
57	`embedding:` block at DashScope / OpenAI / Ollama.
58	- Putting tokens in public places. Team join tokens must never land in a public repo
59	or chat log.

People who liked this also viewed

6 picks

Type