Skip to content
Skills Library
AI & agents · Automation

muster-setup

/muster-setup

Set up or change Muster's workers and routing (~/.muster/config.json). Use for set up muster, add a worker, muster doctor.

Repository
muster ↗
Licence
MIT
Last repository change
13.11.2026

Assessment and security

100/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content35 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
1
Scripts
0
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 13.11.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What it does and when to use it

Set up or change Muster's workers and routing (~/.muster/config.json). Use for set up muster, add a worker, muster doctor.

Original author description

Set up or change Muster's workers and routing (~/.muster/config.json). Use for set up muster, add a worker, muster doctor.

How it works

  1. 1. Detect, don't ask
  2. 2. Propose a config
  3. 3. Ask only what detection cannot answer
  4. 4. Write and check
  5. 5. Hand off
  6. Rules
Original SKILL.md headings · 59 lines, 558 words
SKILL.md preview
---
name: muster-setup
description: Set up or change Muster's workers and routing (~/.muster/config.json). Use for set up muster, add a worker, muster doctor.
---

# Muster setup

Goal: a working `~/.muster/config.json` (or the path in `MUSTER_CONFIG`) that `muster.py doctor` passes, written with as few questions as possible.

The runner is `../muster/muster.py` relative to this skill's base directory. Below, `$MUSTER` means that path.

## 1. Detect, don't ask

```bash
python3 $MUSTER detect
```

It prints JSON:
- `codex_homes`: each folder with a Codex login (`codexHome`, `billing`: `subscription` for a ChatGPT login, `metered` for an API key, `logged_in`).
- `commands`: `codex`, `claude`, and `claude-wrapper` scripts (they set `ANTHROPIC_BASE_URL` and run Claude Code on another backend, such as DeepSeek).
- `config_exists`: when true, read the current config first. You are editing it, not starting over: keep what the user set unless they asked to change it.

Also worth a look, read-only: shell aliases that set `CODEX_HOME` (`grep -n CODEX_HOME ~/.zshrc ~/.bashrc 2>/dev/null`) give each Codex home the name the user already calls it by. Use those names as worker names.

## 2. Propose a config

From what you found, draft workers:
- one `codex` worker per logged-in Codex home: `{ "type": "codex", "codexHome": "<path>" }`
- one `claude` worker per wrapper: `{ "type": "claude", "command": "<path>", "args": ["--dangerously-skip-permissions"], "billing": "metered" }` (drop the flag if the wrapper already passes it)
- a plain `claude` worker only if the user asks for one; it bills the same account this session uses

Then draft `routing` with a `*` fallback, `caps` (`codex` 3, `claude` 5) and `runsDir` (`~/.muster/runs`, or keep the existing one).

## 3. Ask only what detection cannot answer

Use AskUserQuestion, at most one round of up to four questions, each with a recommended default first. Typical:
- Which repos or folders should each worker prefer? (Offer the folders under the user's usual workspace, such as `~/workspace/*`, grouped by prefix.)
- Should any worker be limited to certain repos only (`only`)? Mention that the runner enforces it.
- Parallel caps per account: keep 3 / 5, or change?
- Any rule every worker prompt should carry (`preamble`)? For example, how pull requests are opened in their repos.

If the user already said these things in this conversation, do not ask again.

## 4. Write and check

1. Show the final JSON in one block and say where it goes. If a config exists, show what changes.
2. Write it. Back up an existing file first: `cp ~/.muster/config.json ~/.muster/config.json.bak-$(date +%Y%m%d-%H%M%S)`.
3. Run `python3 $MUSTER doctor`. Fix every line it prints, then run it again until it says `ok`.
4. Optional smoke test, only if the user agrees (it spends a little quota): one task per worker that replies `PONG`, run with `python3 $MUSTER run <worker> <task-dir> /tmp`; report each `status` and `usage`.

## 5. Hand off

Tell the user, in their language:
- the workers and which repos each one takes;
- which limits the runner enforces (`only`, `caps`) and which Claude follows when planning (`routing`, `preamble`);
- how to start: "muster: …" (the pane opens by itself), and `/muster-panel` to reopen it.

## Rules

- Never print the contents of `auth.json`, API key files or environment variables that hold keys. Paths and the `billing` value are enough.
- Do not log in, log out or edit any Codex or Claude profile; if a home is not logged in, tell the user the login command (`CODEX_HOME=<home> codex login`).
- Write only the Muster config (and its backup).

Versions

The database has no other skills named muster-setup.

Database snapshot: 13.11.2026 · package version 6294a3b88d05 (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 100muster
    Run tasks in parallel on Codex and other CLI workers, then verify and report. Use for muster, dispatch, fan out, run in parallel.
  • 90muster-coordination
    Use when this session should coordinate with other coding-agent sessions over the muster bus — registering on the bus, checking your inbox, sending messages or handing tasks to peers, replying on threads, and acting on the notify/nudge/channel wake. Fires when the muster MCP…
  • 100muster
    Dispatch unattended Claude worker sessions over ACP from the main session, then watch them, answer their permission questions, follow up, collect their reports, and clean up. Use when the user asks to dispatch a worker, run a ticket in the background, or delegate work to a headless worker.
  • 100muster
    The ARMADA inspection before sailing. Reviews an open pull request through two independent lenses in parallel — a conventions/correctness code-review pass and a codex-rescue root-cause second opinion — consolidates and dedupes the findings, posts them as inline PR review…
  • 100setup
    Install + verify ax (the agent experience layer). Triggers when the user says "install ax", "set up ax", "ax not found", "ax doctor", "is ax running", "fix ax install", "first-time ax setup", or any setup question about the ax CLI / skills / daemon. Walks the install via the…
  • 87lighthouse
    The ARMADA lighthouse — the fleet's autonomous reconnaissance. Every other ship is reactive (shipwright builds existing issues, muster reviews existing PRs); lighthouse is the one that *proactively finds* work. It surveys the repository for valuable future work — failing/skipped…

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.