muse

v1.4.0

Offload bulk coding work to Muse Code instances in isolated git worktrees, each supervised by an Opus agent that reads the patch, runs the acceptance check and re-prompts until the work is right.

Claude Code1 Skill

By jellologicLicense: MIT0 GitHub starsUpdated 12 hours ago

Directory evidence

Runtimes: Claude Code
Parsed components: 1 skill or MCP entry
Source updated: Sep 23, 2026
Manifest status: Canonical path parsed

The directory validates manifest shape and source location. It does not execute the plugin or provide a security endorsement. Review the indexing methodology →

Install muse for Claude Code

Installs for the current user

claude plugin marketplace add IchenDEV/agent-plugin-mkt
claude plugin marketplace update agent-plugin-marketplace
claude plugin install muse@agent-plugin-marketplace

Paste and run these commands in a terminal with Claude Code. They add and refresh the PluginsMP catalog, then install this plugin.

The installer fetches third-party code from the source repository shown on this page. This directory validates manifest structure and source location, but does not perform a security audit; review the manifest, components, and source before installing.

Get the source manually

$ git clone https://github.com/jellologic/claude-code-muse

Clone the source repository, then follow its setup instructions to add the plugin to a compatible client. The repository root is the plugin root.

Plugin files

muse/

├── .claude-plugin/plugin.json
└── skills/muse-fleet/SKILL.md

Included Skills1

muse-fleetskills/muse-fleet/SKILL.md

Delegate several independent mechanical edits to Muse Code (`muse exec`) workers in isolated git worktrees, each supervised until its patch is right. Use whenever the user names Muse Code, `muse exec`, or a muse run or its results — including asking to put muse on a single coherent change, debugging, or design decisions (the skill tells you to push back), and judging or merging the patches a muse run produced. Also for repetitive refactors or tests across many files, offloading to a cheaper model, fan-out to muse workers, grunt work or "don't burn my tokens on this". Requires the `muse` binary on PATH. Not for parallelism unrelated to muse: human git worktrees, Task-tool/Agent subagents, background jobs.

Plugin manifests1

.claude-plugin/plugin.json

{
  "$schema": "https://json.schemastore.org/claude-code-plugin.json",
  "name": "muse",
  "version": "1.4.0",
  "description": "Offload bulk coding work to Muse Code instances in isolated git worktrees, each supervised by an Opus agent that reads the patch, runs the acceptance check and re-prompts until the work is right.",
  "author": {
    "name": "jellologic",
    "email": "[email protected]",
    "url": "https://github.com/jellologic"
  },
  "homepage": "https://github.com/jellologic/claude-code-muse",
  "repository": "https://github.com/jellologic/claude-code-muse",
  "license": "MIT",
  "keywords": [
    "claude-code",
    "claude-code-plugin",
    "muse",
    "muse-code",
    "delegation",
    "parallel-agents",
    "git-worktree",
    "code-generation",
    "coding-agent",
    "agent-orchestration",
    "cost-optimization",
    "supervision",
    "ai-agents"
  ],
  "displayName": "Muse Code delegation",
  "defaultEnabled": true,
  "workflows": "./workflows",
  "experimental": {
    "monitors": "./monitors/monitors.json"
  },
  "userConfig": {
    "default_effort": {
      "type": "string",
      "title": "Default reasoning effort",
      "description": "Effort every delegated round starts at: none, minimal, low, medium, high, xhigh or max. Teams differ on whether low is right for their codebase. Raise it and every task costs more; lower it and more tasks come back needing a revision round.",
      "default": "low",
      "options": [
        "none",
        "minimal",
        "low",
        "medium",
        "high",
        "xhigh",
        "max"
      ]
    },
    "max_rounds": {
      "type": "number",
      "title": "Maximum revision rounds per task",
      "description": "The runaway-cost breaker. A supervisor gets this many rounds to get a task right before the script refuses rather than letting it loop up a bill.",
      "default": 3,
      "min": 1,
      "max": 10
    },
    "worktree_root": {
      "type": "directory",
      "title": "Where worktrees are created",
      "description": "Leave empty to create them beside the repository, which is the default and keeps them outside it. Set a path if your setup forbids writing next to a checkout. A path INSIDE the repository removes the isolation that makes --yolo defensible. A path containing a single quote is not supported and is refused.",
      "default": ""
    },
    "refuse_on_secrets": {
      "type": "boolean",
      "title": "Refuse to delegate a tree containing a credential",
      "description": "On by default, and worth keeping: contributor-tier models state that content may be used for product improvement, and sending a private key there is not undoable. Turn it off only for a repository whose credential-shaped content is entirely test fixtures.",
      "default": true
    },
    "default_model": {
      "type": "string",
      "title": "Model for delegated work",
      "description": "latest-contributor resolves the newest contributor-tier model from muse's own catalog at run time, so a new generation is used the day it lands. Pin a full-rate id such as muse-spark-1.3 for proprietary or client-confidential code.",
      "default": "latest-contributor"
    }
  }
}