Verify at least 1 issue number is provided
Run gh issue view <number> for each to confirm they exist
When using multiple issues, only parallelize issues that are truly independent — if issue B depends on issue A's output, run B after A, not in parallel
Recommend maximum 4-6 issues for tmux usability

Step 2: Spawn Workers

For each issue:

Create worktree and branch:

git worktree add ../<repo-name>-<issue> -b <git-user>/feature-<issue>

Launch worker pane. The second argument is the initial message sent to the worker's Claude Code session. Use the same model as the coordinator:

${CLAUDE_PLUGIN_ROOT}/skills/parallel-develop/scripts/spawn-worker.sh "../<repo-name>-<issue>" "Resolve the GitHub issue #<issue>" --model <current-model>

To provide extra context, prepend it before the instruction:

${CLAUDE_PLUGIN_ROOT}/skills/parallel-develop/scripts/spawn-worker.sh "../<repo-name>-<issue>" "Focus on the API layer only. Resolve the GitHub issue #<issue>" --model <current-model>

After all workers are spawned:

tmux select-layout main-vertical

Step 3: Output Summary

Provide:

List of worktrees with full paths
Pane navigation: Ctrl+b + arrow keys

Cleanup commands (run in this order):

tmux kill-pane -t <pane-id>
git worktree remove ../<repo-name>-<issue>
git branch -d <git-user>/feature-<issue>

Notes

Workers start in plan mode — they present a plan for approval before implementing
Switch to a worker pane with Ctrl+b + arrow keys to intervene if needed
If a pane gets stuck, use Ctrl+c to interrupt
Resolve <git-user> with git config user.name | tr ' ' '-' | tr '[:upper:]' '[:lower:]'
Resolve <repo-name> from git remote get-url origin