Ultrawork Exec Command

Execute a plan created by /ultrawork-plan. Includes automatic retry loop for failed tasks and verification.

---

Overview

/ultrawork-exec
    ↓
Load Session → Execute Tasks (waves) → Verify → Retry if failed
    ↓
Loop until: PASS or max_iterations reached

---

Delegation Rules (MANDATORY)

The orchestrator MUST delegate work to sub-agents. Direct execution is prohibited.

Phase	Delegation	Direct Execution
Execution	ALWAYS via Task(subagent_type="ultrawork:worker")	NEVER
Verification	ALWAYS via Task(subagent_type="ultrawork:verifier")	NEVER

Exception: User explicitly requests direct execution (e.g., "run this directly", "execute without agent").

---

Sub-agent Execution

Sub-agents can be run in foreground (default) or background mode. Choose based on the situation:

Mode	When to Use
Foreground	Sequential tasks, need result immediately
Background	Parallel execution with worker pool limits

Parallel execution: Call multiple Tasks in a single message for automatic parallelization.

---

Session ID Handling (CRITICAL)

All scripts require --session <id> flag.

Where to get SESSION_ID

Look for this message in system-reminder (provided by SessionStart hook):

CLAUDE_SESSION_ID: 37b6a60f-8e3e-4631-8f62-8eaf3d235642
Use this when calling ultrawork scripts: --session 37b6a60f-8e3e-4631-8f62-8eaf3d235642

IMPORTANT: You MUST extract the actual UUID value and use it directly. DO NOT use placeholder strings like {SESSION_ID} or $SESSION_ID.

Correct usage example

If the hook says CLAUDE_SESSION_ID: 37b6a60f-8e3e-4631-8f62-8eaf3d235642, then:

# ✅ CORRECT - use the actual value
node "${CLAUDE_PLUGIN_ROOT}/src/scripts/session-get.js" --session 37b6a60f-8e3e-4631-8f62-8eaf3d235642 --field phase

# ❌ WRONG - do not use placeholders
node "${CLAUDE_PLUGIN_ROOT}/src/scripts/session-get.js" --session {SESSION_ID} --field phase

Script calls with SESSION_ID

All scripts accept session ID directly and derive the directory internally:

node "${CLAUDE_PLUGIN_ROOT}/src/scripts/session-get.js" --session <SESSION_ID> --field phase
node "${CLAUDE_PLUGIN_ROOT}/src/scripts/session-update.js" --session <SESSION_ID> --phase EXECUTION
node "${CLAUDE_PLUGIN_ROOT}/src/scripts/task-list.js" --session <SESSION_ID> --format json

---

Step 1: Load Session

Find and load the existing session:

# Get SESSION_ID from hook output (see "Session ID Handling" section)
# Get session directory via script
SESSION_DIR=$(node "${CLAUDE_PLUGIN_ROOT}/src/scripts/session-get.js" --session {SESSION_ID} --dir)

# Verify session exists and has tasks
node "${CLAUDE_PLUGIN_ROOT}/src/scripts/task-list.js" --session {SESSION_ID}

Read session state:

# SESSION_ID from hook output
# Get session data via script
node "${CLAUDE_PLUGIN_ROOT}/src/scripts/session-get.js" --session {SESSION_ID}                    # Full JSON
node "${CLAUDE_PLUGIN_ROOT}/src/scripts/session-get.js" --session {SESSION_ID} --field phase      # Specific field
node "${CLAUDE_PLUGIN_ROOT}/src/scripts/session-get.js" --session {SESSION_ID} --field goal

# Get specific options
goal=$(node "${CLAUDE_PLUGIN_ROOT}/src/scripts/session-get.js" --session {SESSION_ID} --field goal)
phase=$(node "${CLAUDE_PLUGIN_ROOT}/src/scripts/session-get.js" --session {SESSION_ID} --field phase)

Validate session is ready for execution:

Phase	Action
PLANNING_COMPLETE	Ready to execute
EXECUTION	Resume execution (check task states)
VERIFICATION	Resume verification
COMPLETE	Already done, report success
FAILED	Max iterations reached, report failure
CANCELLED	Exit cleanly

---

Step 2: Show Execution Plan

Display plan summary before starting:

## Starting Execution

**Goal:** {goal}
**Session:** {session_id}
**Iteration:** {iteration}/{max_iterations}
**Tasks:** {count}

### Task Queue

| ID     | Task         | Status  | Complexity | Model  |
| ------ | ------------ | ------- | ---------- | ------ |
| 1      | Setup schema | pending | standard   | sonnet |
| 2      | Build API    | pending | complex    | opus   |
| verify | Verification | pending | complex    | opus   |

### Execution Order
1. [READY] 1 - no dependencies
2. [BLOCKED] 2 - depends on 1
3. [BLOCKED] verify - depends on all

Starting workers...

---

Step 3: Execution Loop (Main Retry Loop)

The execution loop runs until success or max_iterations reached.

while iteration <= max_iterations:
    # Update phase and iteration
    Bash(f'node "{CLAUDE_PLUGIN_ROOT}/src/scripts/session-update.js" --session {SESSION_ID} --phase EXECUTION --iteration {iteration}')

    print(f"## Iteration {iteration}/{max_iterations}")

    # Run execution phase
    execution_result = run_execution_phase(SESSION_ID, max_workers)

    if execution_result == "CANCELLED":
        return

    # Skip verify if requested
    if skip_verify:
        Bash(f'node "{CLAUDE_PLUGIN_ROOT}/src/scripts/session-update.js" --session {SESSION_ID} --phase COMPLETE')
        print("## Execution Complete (verification skipped)")
        return

    # Run verification phase
    verification_result = run_verification_phase(SESSION_ID)

    if verification_result == "CANCELLED":
        return
    elif verification_result == "PASS":
        Bash(f'node "{CLAUDE_PLUGIN_ROOT}/src/scripts/session-update.js" --session {SESSION_ID} --phase COMPLETE')
        print("## Execution Complete - All criteria verified")
        return
    else:
        # FAIL - retry if iterations remain
        if iteration < max_iterations:
            print(f"## Verification Failed - Retrying ({iteration + 1}/{max_iterations})")
            reset_failed_tasks(SESSION_ID)
            iteration += 1
        else:
            Bash(f'node "{CLAUDE_PLUGIN_ROOT}/src/scripts/session-update.js" --session {SESSION_ID} --phase FAILED')
            print("## Execution Failed - Max iterations reached")
            return

---

Step 4: Execution Phase Implementation

def run_execution_phase(SESSION_ID, max_workers):
    # Get session_dir via: Bash(f'node "{CLAUDE_PLUGIN_ROOT}/src/scripts/session-get.js" --session {SESSION_ID} --dir')

    while True:
        # Get current task states
        tasks_output = Bash(f'node "{CLAUDE_PLUGIN_ROOT}/src/scripts/task-list.js" --session {SESSION_ID} --format json')
        tasks = json.loads(tasks_output.output)

        # Categorize tasks (exclude verify task for now)
        non_verify_tasks = [t for t in tasks if t["id"] != "verify"]
        unblocked = [t for t in non_verify_tasks if t["status"] == "pending" and all_deps_complete(t, tasks)]
        all_done = all(t["status"] == "resolved" for t in non_verify_tasks)

        if all_done:
            return "DONE"  # Move to verification

        # Spawn workers for unblocked tasks
        # Call multiple Tasks in single message = automatic parallel execution
        batch = unblocked[:max_workers] if max_workers > 0 else unblocked
        for task in batch:
            # Mark task as in_progress
            Bash(f'node "{CLAUDE_PLUGIN_ROOT}/src/scripts/task-update.js" --session {SESSION_ID} --id {task["id"]} --status in_progress')

            model = "opus" if task["complexity"] == "complex" else "sonnet"
            Task(
                subagent_type="ultrawork:worker:worker",
                model=model,
                prompt=f"""
SESSION_ID: {SESSION_ID}
TASK_ID: {task["id"]}

TASK: {task["subject"]}
{task["description"]}

SUCCESS CRITERIA:
{task["criteria"]}
"""
            )
            print(f"→ Started: {task['id']} - {task['subject']}")
        # All workers in batch complete before next iteration

---

Step 5: Verification Phase Implementation

def run_verification_phase(SESSION_ID):
    # Get session_dir via: Bash(f'node "{CLAUDE_PLUGIN_ROOT}/src/scripts/session-get.js" --session {SESSION_ID} --dir')

    # Update phase
    Bash(f'node "{CLAUDE_PLUGIN_ROOT}/src/scripts/session-update.js" --session {SESSION_ID} --phase VERIFICATION')

    print("## Running Verification...")

    # Spawn verifier (foreground - waits for completion)
    result = Task(
        subagent_type="ultrawork:verifier:verifier",
        model="opus",
        prompt=f"""
SESSION_ID: {SESSION_ID}

Verify all success criteria are met with evidence.
Check for blocked patterns.
Run final tests.

Return: PASS or FAIL with details
"""
    )

    # Parse verifier output
    if "PASS" in result.output:
        return "PASS"
    else:
        return "FAIL"

---

Step 6: Reset Failed Tasks for Retry

def reset_failed_tasks(SESSION_ID):
    """Reset failed tasks for retry iteration"""

    tasks_output = Bash(f'node "{CLAUDE_PLUGIN_ROOT}/src/scripts/task-list.js" --session {SESSION_ID} --format json')
    tasks = json.loads(tasks_output.output)

    for task in tasks:
        if task["status"] == "failed":
            # Reset to pending
            Bash(f'node "{CLAUDE_PLUGIN_ROOT}/src/scripts/task-update.js" --session {SESSION_ID} --id {task["id"]} --status pending')

            # Increment retry count
            retry_count = task.get("retry_count", 0) + 1
            Bash(f'node "{CLAUDE_PLUGIN_ROOT}/src/scripts/task-update.js" --session {SESSION_ID} --id {task["id"]} --retry-count {retry_count}')

            print(f"↻ Reset for retry: {task['id']} (attempt {retry_count + 1})")

    # Also reset verify task
    Bash(f'node "{CLAUDE_PLUGIN_ROOT}/src/scripts/task-update.js" --session {SESSION_ID} --id verify --status pending')

---

Step 7: Progress Reporting

Between poll iterations, report progress:

## Progress Update

**Iteration:** 2/5
**Phase:** EXECUTION

| Task   | Status      | Model  |
| ------ | ----------- | ------ |
| 1      | ✓ resolved  | sonnet |
| 2      | → running   | opus   |
| 3      | ⏳ blocked  | sonnet |
| verify | ⏳ pending  | opus   |

Active workers: 1

---

Step 8: Completion

On Success (PASS)

## Execution Complete

**Goal:** {goal}
**Iterations:** {iteration}/{max_iterations}
**Result:** ✓ PASS

### Summary
- All {task_count} tasks completed
- All success criteria verified
- No blocked patterns detected

### Evidence
{summary from verifier}

Session ID: {session_id}

On Failure (max iterations reached)

## Execution Failed

**Goal:** {goal}
**Iterations:** {max_iterations}/{max_iterations} (max reached)
**Result:** ✗ FAIL

### Failed Tasks
| ID | Task | Reason |
|----|------|--------|
| 2  | Build API | Tests failing |

### Verification Issues
{details from last verifier run}

### Next Steps
1. Review failed task output: ~/.claude/ultrawork/sessions/{session_id}/tasks/2.json
2. Fix issues manually
3. Run `/ultrawork-exec` again (resets iteration counter)

Session ID: {session_id}

---

Error Handling

Worker Failures

• Mark task as failed in session.json
• Continue other independent tasks
• Retry on next iteration

Retry Strategy

Attempt	Model Escalation	Notes
1	sonnet/opus (default)	Normal execution
2	Same model	Retry with same model
3+	Consider opus	Escalate if standard task keeps failing

Graceful Degradation

• If a non-critical task fails repeatedly, consider marking as skipped
• Critical tasks (verify) must pass

---

Options Reference

Option	Description
--session <id>	Session ID to execute
--max-iterations N	Override max retry iterations (default: 5)
--skip-verify	Skip verification phase

---

Directory Structure

Get session directory: node "${CLAUDE_PLUGIN_ROOT}/src/scripts/session-get.js" --session {SESSION_ID} --dir

$SESSION_DIR/
├── session.json        # Session metadata with iteration state
├── context.json        # Explorer summaries
├── design.md           # Design document
├── exploration/        # Exploration files
└── tasks/              # Task files with status and retry_count
    ├── 1.json          # { status: "resolved", retry_count: 0 }
    ├── 2.json          # { status: "failed", retry_count: 2 }
    └── verify.json     # { status: "pending" }

---

Zero Tolerance Rules

Before ANY completion claim:

• No blocked phrases ("should work", "basic implementation")
• Evidence exists for all criteria
• All tasks resolved
• Verifier passed (unless --skip-verify)
• Retry loop exhausted only after genuine attempts