agentleFS
Tony363Tony363/SuperClaude/.claude/skills/sc-worktree/SKILL.md

sc-worktree

Skill19 starsChanged 8 months agoMITView on GitHub

Git worktree management for multi-agent parallel development. Use when spawning multiple agents to work on isolated features simultaneously.

  • Commits and pushes

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 10 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈10

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Distinctive:

    Cleanup: - Remove merged/stale worktrees

    line 33
  2. Distinctive:

    When spawning a subagent to work in a worktree:

    line 138
  3. Distinctive:

    Before creating a worktree, the skill validates:

    line 161
  4. Distinctive:

    No conflicting remote branch

    line 169
  5. Distinctive:

    Before proposing, the skill checks for conflicts:

    line 176
  6. Distinctive:

    Read: - Check worktree status and logs

    line 311
  7. Distinctive:

    No automatic quality scoring (use CI/tests)

    line 326
  8. Distinctive:

    No lease tracking (periodic cleanup instead)

    line 327
  9. Distinctive:

    No sophisticated conflict resolution (human review)

    line 328
  10. Distinctive:

    No auto-merge to main (PR-based workflow)

    line 329

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Worktree Management Skill

Git worktree isolation for parallel agent development with simple, safe workflows.

Quick Start

# Create isolated worktree for a task
/sc:worktree create feature-auth

# List all active worktrees
/sc:worktree list

# Propose changes for human review
/sc:worktree propose feature-auth

# Clean up stale worktrees
/sc:worktree cleanup

Behavioral Flow

  1. Create - Initialize isolated worktree with unique naming
  2. Validate - Preflight checks (clean repo, no conflicts)
  3. Assign - Spawn subagent with CWD set to worktree
  4. Work - Agent operates in isolation, commits to branch
  5. Propose - Push branch and create PR for human review
  6. Cleanup - Remove merged/stale worktrees

Commands

CommandPurposeGit Operations
create <task-id>Create isolated worktreegit worktree add -b wt/<task>
listShow all active worktreesgit worktree list
status <task-id>Check worktree stategit status, git log
cleanup [--all]Remove stale worktreesgit worktree remove, prune
propose <task-id>Create PR for human reviewgit push, gh pr create

Flags

FlagTypeDefaultDescription
--commitsbooltrueAgent commits directly to branch
--patchesboolfalseAgent produces patch files only
--run-testsboolfalseRun tests before proposing
--forceboolfalseOverride existing worktree/branch
--basestringmainBase branch for worktree

Evidence Requirements

This skill does NOT require hard evidence. Worktree operations are self-documenting through:

  • Git worktree list output
  • Branch history
  • PR creation confirmation

Naming Convention

Worktrees use deterministic naming to prevent collisions:

Path:   .worktrees/<task-id>/
Branch: wt/<task-id>

Example:

Path:   .worktrees/feature-auth/
Branch: wt/feature-auth

Operations

Create Worktree

/sc:worktree create feature-auth

# Executes:
# 1. mkdir -p .worktrees
# 2. git worktree add -b wt/feature-auth .worktrees/feature-auth
# 3. Returns worktree path for subagent CWD

List Worktrees

/sc:worktree list

# Shows:
# - Active worktrees
# - Associated branches
# - Last commit info

Check Status

/sc:worktree status feature-auth

# Shows:
# - Uncommitted changes
# - Commits ahead of base branch
# - Conflict warnings

Propose Changes

/sc:worktree propose feature-auth

# Executes:
# 1. git push -u origin wt/feature-auth
# 2. gh pr create --title "feat: feature-auth" --body "..."
# 3. Returns PR URL for human review

Cleanup Worktrees

/sc:worktree cleanup          # Remove merged worktrees
/sc:worktree cleanup --all    # Remove ALL worktrees
/sc:worktree cleanup feature-auth  # Remove specific worktree

# Executes:
# 1. git worktree remove .worktrees/<task-id>
# 2. git branch -d wt/<task-id>
# 3. git worktree prune

Subagent Integration

When spawning a subagent to work in a worktree:

# 1. Create worktree
/sc:worktree create feature-auth

# 2. Spawn subagent with CWD
Task(
    prompt="Implement JWT authentication",
    subagent_type="general-purpose",
    # CWD implicitly set to .worktrees/feature-auth/
)

# 3. Agent works in isolation
# - All file operations scoped to worktree
# - Commits go to wt/feature-auth branch

# 4. Propose for review
/sc:worktree propose feature-auth

Preflight Checks

Before creating a worktree, the skill validates:

  1. Repository State
    • Repo is clean OR --force flag provided
    • Main branch is up to date
  2. Branch Availability
    • wt/<task-id> branch doesn't exist OR explicit reuse
    • No conflicting remote branch
  3. Path Availability
    • .worktrees/<task-id> doesn't exist OR --force flag

Conflict Detection

Before proposing, the skill checks for conflicts:

# Fetch latest changes
git fetch origin main

# Check if worktree is behind
git merge-base --is-ancestor origin/main HEAD

# Warn but don't block if behind
# Human reviewer handles merge strategy

Examples

Single Agent Workflow

# Create worktree
/sc:worktree create add-logging

# Work in worktree (agent operates here)
# ... agent makes changes, commits ...

# Propose changes
/sc:worktree propose add-logging

# Human reviews PR, merges

# Cleanup
/sc:worktree cleanup add-logging

Parallel Multi-Agent Workflow

# Orchestrator creates multiple worktrees
/sc:worktree create feature-auth
/sc:worktree create feature-logging
/sc:worktree create feature-metrics

# Spawn agents in parallel (each with own worktree)
# Agent A -> .worktrees/feature-auth/
# Agent B -> .worktrees/feature-logging/
# Agent C -> .worktrees/feature-metrics/

# Propose all changes
/sc:worktree propose feature-auth
/sc:worktree propose feature-logging
/sc:worktree propose feature-metrics

# Human reviews 3 PRs, merges sequentially

# Cleanup all
/sc:worktree cleanup --all

Patch Mode (No Direct Commits)

/sc:worktree create refactor-api --patches

# Agent produces .patch files instead of commits
# Orchestrator reviews patches before applying

/sc:worktree propose refactor-api  # Creates PR from patches

MCP Integration

PAL MCP (Validation & Review)

ToolWhen to UsePurpose
mcp__pal__precommitBefore proposeValidate changes in worktree
mcp__pal__codereviewBefore proposeReview worktree changes
mcp__pal__consensusConflict resolutionMulti-model merge strategy

PAL Usage Patterns

# Validate worktree changes before proposing
mcp__pal__precommit(
    path=".worktrees/feature-auth",
    step="Validating worktree changes",
    findings="Security, test coverage, completeness",
    compare_to="main"
)

# Review worktree branch
mcp__pal__codereview(
    review_type="full",
    step="Reviewing feature-auth worktree",
    relevant_files=[".worktrees/feature-auth/src/auth.py"]
)

Rube MCP (Automation & Notifications)

ToolWhen to UsePurpose
mcp__rube__RUBE_SEARCH_TOOLSGitHub integrationFind PR tools
mcp__rube__RUBE_MULTI_EXECUTE_TOOLPropose commandCreate PR, notify team

Rube Usage Patterns

# Create PR and notify team
mcp__rube__RUBE_MULTI_EXECUTE_TOOL(tools=[
    {"tool_slug": "GITHUB_CREATE_PULL_REQUEST", "arguments": {
        "repo": "myapp",
        "title": "feat: feature-auth from worktree",
        "body": "## Worktree: feature-auth\nCreated via /sc:worktree",
        "base": "main",
        "head": "wt/feature-auth"
    }},
    {"tool_slug": "SLACK_SEND_MESSAGE", "arguments": {
        "channel": "#pull-requests",
        "text": "New PR from worktree: wt/feature-auth"
    }}
])

Flags (Extended)

FlagTypeDefaultDescription
--pal-reviewboolfalseUse PAL codereview before propose
--create-prbooltrueCreate GitHub PR via Rube
--notifystring-Notify via Rube (slack, teams)
--draftboolfalseCreate draft PR

Tool Coordination

  • Bash - Git worktree commands, directory operations
  • Read - Check worktree status and logs
  • Task - Spawn subagents with worktree CWD
  • PAL MCP - Pre-propose validation, code review
  • Rube MCP - PR creation, team notifications

Safety Guarantees

  1. Isolation - Each worktree has separate working directory
  2. No Auto-Merge - Human review required for all merges
  3. Unique Naming - Deterministic paths prevent collisions
  4. Crash-Safe - cleanup --all removes stale worktrees
  5. Warning-Only Conflicts - Conflicts warn but don't block

Limitations

  • No automatic quality scoring (use CI/tests)
  • No lease tracking (periodic cleanup instead)
  • No sophisticated conflict resolution (human review)
  • No auto-merge to main (PR-based workflow)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.