.claude/skills/worktree/SKILL.md
Create and manage git worktrees for parallel agent sessions
npx skillsauth add DavidROliverBA/ArchitectKB worktreeInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Manage git worktrees for parallel Claude Code sessions and subagent dispatch.
| Command | Usage | Purpose |
|---------|-------|---------|
| /worktree create <slug> | Create a new worktree | Isolated workspace for parallel work |
| /worktree list | List active worktrees | See what's running |
| /worktree merge <slug> | Merge and clean up | Bring work back to main |
| /worktree cleanup <slug> | Remove without merging | Abandon a worktree |
All commands delegate to .claude/scripts/worktree-manager.sh:
.claude/scripts/worktree-manager.sh <command> [slug]
IMPORTANT: The script operates on the vault root. Always run from the vault directory. Use dangerouslyDisableSandbox: true because worktrees are created outside the vault directory (sibling path).
meeting-notes-batch, architecture-reviewworktree/<slug>../<vault>-worktrees/<slug>/After creating a worktree, dispatch subagents with the worktree path as their working directory:
1. Create worktree: /worktree create meeting-notes-batch
2. Note the path from output (e.g. /Users/.../BA-DavidOliver-ObsidianVault-worktrees/meeting-notes-batch)
3. Dispatch subagent with Task tool, including in the prompt:
"Work in /Users/.../BA-DavidOliver-ObsidianVault-worktrees/meeting-notes-batch"
4. When subagent completes: /worktree merge meeting-notes-batch
For separate Claude Code terminals working in parallel:
/worktree create feature-alpha → opens terminal in worktree path/worktree create feature-beta → opens terminal in worktree path/worktree merge feature-alpha (from the main vault session)worktree/<slug> — visually distinct in git log--no-ff for clean merge commits.claude/ and .mcp.json are copied to each worktree.obsidian/ — worktrees are agent workspaces, not Obsidian vaults../<vault>-worktrees/<slug>/ (outside vault, invisible to Obsidian)| Issue | Fix |
|-------|-----|
| "slug required" | Provide a slug: /worktree create my-task |
| Merge conflicts | Resolve manually in main, then git worktree remove |
| Sandbox blocks worktree creation | Use dangerouslyDisableSandbox: true |
| Worktree already exists | Idempotent — reuses existing worktree |
tools
--- context: fork --- # /youtube Save a YouTube video as both a Weblink (quick reference) and a detailed Page (full analysis). ## Usage ``` /youtube <url> /youtube <url> <optional title override> ``` ## Examples ``` /youtube https://www.youtube.com/watch?v=0TpON5T-Sw4 /youtube https://youtu.be/abc123 AWS re:Invent Keynote ``` ## Prerequisites This skill uses the MCP Docker YouTube tools: - `mcp__MCP_DOCKER__get_video_info` - Video metadata - `mcp__MCP_DOCKER__get_transcript` - Full trans
testing
--- context: fork --- # /wipe Generate a context handoff summary, clear the session, and resume in a fresh conversation. Detects environment and provides automated (tmux) or manual workflow. ## Usage ``` /wipe /wipe quick # Minimal handoff, just essentials /wipe detailed # Comprehensive handoff with full context ``` ## Instructions When the user invokes `/wipe`: ### Phase 1: Detect Environment First, check the terminal environment: ```bash echo "Environment Detection:"
data-ai
--- context: fork --- # /weekly-summary Generate comprehensive weekly summary from daily notes, meetings, tasks, and project updates using parallel sub-agents. ## Usage ``` /weekly-summary /weekly-summary --last-week /weekly-summary --from 2026-01-01 --to 2026-01-07 /weekly-summary --output page # Create Page note instead of just outputting ``` ## Instructions This skill uses **5 parallel sub-agents** to gather data concurrently from different vault areas, then synthesizes a comprehensi
development
--- context: fork --- # /weblink Save a URL as a weblink note with analysis and summary. ## Usage ``` /weblink <url> /weblink <url> <optional title> ``` ## Examples ``` /weblink https://github.com/anthropics/claude-code /weblink https://aws.amazon.com/bedrock/ AWS Bedrock Service ``` ## Instructions 1. **Fetch and analyse the URL**: - Use WebFetch to retrieve the page content - Extract: title, author, source/domain, main content - **ALWAYS provide analysis** - this is mandatory,