Power Resume Invoker

Resume interrupted or paused orchestration workflows from their last checkpoint, restoring full context and continuing execution without repeating completed steps.

When to Use This Command

Run this command when...

• A long-running orchestration workflow was interrupted by a timeout, crash, or manual pause and needs to continue
• You want to pick up where a previous Claude session left off without re-executing completed tasks
• An error caused a workflow to stop mid-execution and you need to retry from the failure point
• You are returning to a multi-session project and want to restore the full orchestration context
• A background task completed and the remaining workflow steps need to be triggered

Quick Start

Copy
# .claude/commands/power-resume-invoker.md
---
name: Power Resume Invoker
description: Resume orchestration from last checkpoint with full context restore
command: true
---
Resume workflow: $ARGUMENTS

1. Locate the last checkpoint or saved state
2. Restore context: completed tasks, pending tasks, variables
3. Validate current state matches expectations
4. Continue execution from the next pending step

Copy
# Invoke the command
claude "/power-resume-invoker"

# Expected output
# > Scanning for orchestration state...
# > Found: task-orchestration/2026-03-15/
# >   Completed: TASK-001 through TASK-008 (8 tasks)
# >   Failed: TASK-009 (timeout on API integration test)
# >   Pending: TASK-010 through TASK-014 (5 tasks)
# > Restoring context from checkpoint...
# > Retrying TASK-009 with extended timeout...
# > Resuming workflow from TASK-009

Core Concepts

Concept	Description
Checkpoint Detection	Scans orchestration directories for saved state files and progress markers
Context Restoration	Rebuilds the full variable state, completed work list, and dependency graph
Failure Recovery	Identifies why execution stopped and applies retry logic or workarounds
Skip Completed	Never re-executes tasks that already succeeded, saving time and avoiding side effects
State Validation	Verifies that the project state on disk matches expected checkpoint state before resuming

Resume Invoker Flow
====================

  Scan State --> Identify Checkpoint --> Validate --> Resume
      |                  |                   |           |
  Find task-        Load completed      Compare       Continue
  orchestration/    and pending         disk state    from next
  directories       task lists          vs saved      pending task
      |                  |                   |           |
  Locate latest     Restore context     Flag any      Execute
  state files       and variables       drift or      remaining
                                        conflicts     tasks

Configuration

Parameter	Description	Default	Example	Required
$ARGUMENTS	Workflow ID or directory to resume from	auto-detect latest	"2026-03-15"	No
retry_failed	Automatically retry the task that caused the interruption	true	false	No
max_retries	Maximum retry attempts for a failed task before skipping	2	5	No
restore_env	Restore environment variables from checkpoint	true	false	No
skip_validation	Skip state validation check before resuming	false	true	No

Best Practices

1. Let auto-detection find the latest state -- The invoker scans for the most recent orchestration directory by default. Only specify a workflow ID when resuming an older session.

2. Review the failure reason before retrying -- When a task failed, understand why before automatically retrying. Timeouts may need extended limits; auth errors need credential updates.

3. Validate state after long gaps -- If days have passed since the interruption, files may have changed. Keep skip_validation: false to catch drift between saved state and current disk state.

4. Commit work before resuming -- If you made manual changes since the interruption, commit them so the resume has a clean baseline and can properly attribute changes.

5. Use for multi-session projects -- Break large projects into sessions. End each session cleanly, then use the resume invoker next time to pick up with full context.

Common Issues

Checkpoint state not found: If orchestration state files were deleted or the directory structure changed, the invoker cannot resume. Ensure task-orchestration directories are preserved between sessions.

State drift causes validation failure: When files were modified manually after the checkpoint, validation detects mismatches. Use skip_validation: true cautiously, or reconcile changes before resuming.

Infinite retry loops on persistent failures: A task that fails for structural reasons like a missing API or incorrect schema will fail on every retry. Set max_retries appropriately and add --skip TASK-ID to bypass permanently broken tasks.