Implement

If worktree missing, create it:

git worktree add ../{feature-name}-wt -b feat/{feature-name}

Phase 2: Parse Tasks

Extract ordered task list from plan.md. Build execution sequence respecting dependencies.

Load docs/coding-principles.md — include its key rules as context when delegating to workers.

Phase 3: Execute Tasks

For each task in order:

Option A — Docker isolated (preferred):

docker run --rm --network none \
  -v "{worktree}:/workspace" \
  -w /workspace {base-image} \
  codex exec -m gpt-5.3-codex -c model_reasoning_effort=xhigh "{task-prompt}"

Option B — Direct execution (fallback): If Docker unavailable, execute in worktree directly:

• Spawn worker agent for each task
• Worker operates within worktree directory
• guard.sh restricts file access to worktree

Phase 4: Per-Task Verification

After each task:

1. Run acceptance criteria checks
2. Commit changes in worktree:

   cd {worktree} && git add -A && git commit -m "feat({feature}): {task title}"
   

3. Log result to .dev/state.json

Phase 5: Completion

After all tasks:

1. Push feature branch: git push -u origin feat/{feature-name}
2. Update .dev/state.json with completion status
3. Send notification via scripts/notify.sh

Orchestrator Rules

• You are the Orchestrator — delegate to workers, don't implement directly
• Each worker gets ONE task — no multi-task workers
• Workers MUST NOT use Task tool (no sub-delegation)
• If a task fails, stop and report — don't skip ahead
• Commit after each successful task (atomic commits)