coalesce-labs/setup-orchestrate

Setup Orchestrate

Create an orchestrator worktree and output a single copy-paste command to launch the orchestration run. No questions — parse the input, create the worktree, print the command.

Input Format

/catalyst-dev:setup-orchestrate <ticket-ids...>
/catalyst-dev:setup-orchestrate --cycle current
/catalyst-dev:setup-orchestrate --project "Project Name"
/catalyst-dev:setup-orchestrate --auto N

The input MUST be one of:

• Ticket IDs: space-separated (e.g., ADV-214 ADV-215 ADV-208)
• Cycle flag: --cycle current
• Project flag: --project "Project Name"
• Auto flag: --auto N — orchestrate skill will pick top N Todo tickets

If no input is provided, stop and tell the user:

Usage: /catalyst-dev:setup-orchestrate <ticket-ids...>
       /catalyst-dev:setup-orchestrate --cycle current
       /catalyst-dev:setup-orchestrate --project "Project Name"
       /catalyst-dev:setup-orchestrate --auto N

Do NOT analyze tickets, suggest sequencing, research ticket details, or do any work beyond creating the worktree. Wave planning and dependency analysis is the orchestrate skill's job.

Process

Execute all steps without asking questions. No confirmations, no menus, no options.

Step 1: Parse Input

Determine which mode the user invoked:

• If the input contains --cycle, extract the value (e.g., current)
• If the input contains --project, extract the quoted value (e.g., "Project Name")
• If the input contains --auto, extract the integer N that follows (e.g., 5)
• Otherwise, treat all tokens as space-separated ticket IDs

Step 2: Call setup-orchestrator.sh

Build the flag string from the parsed input and call the standalone script:

# For ticket IDs:
"${CLAUDE_PLUGIN_ROOT}/scripts/setup-orchestrator.sh" --tickets "<ticket-ids>"

# For cycle mode:
"${CLAUDE_PLUGIN_ROOT}/scripts/setup-orchestrator.sh" --cycle current

# For project mode:
"${CLAUDE_PLUGIN_ROOT}/scripts/setup-orchestrator.sh" --project "<project-name>"

# For auto mode:
"${CLAUDE_PLUGIN_ROOT}/scripts/setup-orchestrator.sh" --auto <N>

The script handles everything: validation, config reading, state init, worktree creation.

Step 3: Format Output

The script prints a formatted output block with the worktree path and launch commands. Display the script's output directly — it already includes the ═══ banner with dry-run and full-run commands, plus the machine-readable WORKTREE_PATH=... line.

Rules

• No questions. Parse input, execute, print command. The entire skill should complete in one turn with zero user interaction.
• No ticket analysis. Do not fetch ticket details, suggest sequencing, or build wave plans. That is Phase 1 of the orchestrate skill.
• No worktree option. Always refuse if not on main repo. Do not offer alternatives.
• No custom names. Always auto-generate orch-YYYY-MM-DD[-N].
• Always show both commands — dry run first, full run second.
• Do NOT run the orchestrate skill — the user runs it from the new worktree in a new terminal.