project-standalo-sonic-cloud/.claude/commands/workflow/resume.md

description	allowed-tools
Resume an interrupted workflow from saved state	Read, Write, Edit, Bash, AskUserQuestion, TodoWrite

Workflow Orchestrator - Resume

Resume a previously interrupted or paused workflow.

EXECUTION PROTOCOL

Step 1: Load Workflow State

Read .workflow/current.yml:

• If not found: Report "No workflow to resume" and exit
• If found: Load state and continue

Step 2: Display Resume Summary

╔══════════════════════════════════════════════════════════════╗
║              🔄 RESUMING WORKFLOW                             ║
╠══════════════════════════════════════════════════════════════╣
║ Workflow ID: <id>                                             ║
║ Feature: <feature>                                            ║
║ Started: <started_at>                                         ║
║ Last Updated: <updated_at>                                    ║
╠══════════════════════════════════════════════════════════════╣
║ CURRENT STATE                                                 ║
║   Phase: <current_phase>                                      ║
║   Resume Point: <resume_point.action>                         ║
╠══════════════════════════════════════════════════════════════╣
║ PROGRESS                                                      ║
║   Entities Designed:    <progress.entities_designed>          ║
║   Tasks Created:        <progress.tasks_created>              ║
║   Tasks Implemented:    <progress.tasks_implemented>          ║
║   Tasks Reviewed:       <progress.tasks_reviewed>             ║
║   Tasks Completed:      <progress.tasks_completed>            ║
╠══════════════════════════════════════════════════════════════╣
║ LAST ERROR (if any): <last_error>                             ║
╚══════════════════════════════════════════════════════════════╝

Step 3: Confirm Resume

Ask user:

• Option 1: "Continue - Resume from current point"
• Option 2: "Restart Phase - Redo current phase from beginning"
• Option 3: "Abort - Cancel workflow entirely"

Step 4: Resume Based on Phase

INITIALIZING: → Continue to DESIGNING phase

DESIGNING: → Continue architect work → Resume creating entities/tasks

AWAITING_DESIGN_APPROVAL: → Present design summary again → Ask for approval

DESIGN_APPROVED: → Continue to IMPLEMENTING phase

DESIGN_REJECTED: → Show rejection reason → Return to DESIGNING with feedback

IMPLEMENTING: → Find incomplete tasks → Continue implementation from next pending task

REVIEWING: → Find tasks awaiting review → Continue review process

SECURITY_REVIEW: → Continue security scanning → Run: python3 skills/guardrail-orchestrator/scripts/security_scan.py --project-dir . --severity HIGH → Run: python3 skills/guardrail-orchestrator/scripts/validate_api_contract.py --project-dir . → If passed: Transition to AWAITING_IMPL_APPROVAL → If critical issues: Return to IMPLEMENTING with security feedback

AWAITING_IMPL_APPROVAL: → Present implementation summary again → Ask for approval

IMPL_APPROVED: → Continue to COMPLETING phase

IMPL_REJECTED: → Show rejection reason → Return to IMPLEMENTING with feedback

COMPLETING: → Continue marking tasks complete

PAUSED: → Resume from resume_point.phase

FAILED: → Show error details → Ask user how to proceed:

• Retry failed operation
• Skip and continue
• Abort workflow

Step 5: Continue Workflow

Execute remaining phases following /workflow:spawn protocol.

TASK-LEVEL RESUME

If resuming during IMPLEMENTING phase:

1. Identify incomplete tasks:

   # Resume from first task not in 'completed' or 'approved'
   resume_task: tasks.pending[0] || tasks.in_progress[0] || tasks.review[0]
   

2. Skip completed work:

   • Don't recreate files that exist and are valid
   • Don't re-run validations that passed

3. Continue from failure point:

   • If task failed mid-implementation, restart that task
   • If validation failed, show error and retry

STATE RECOVERY

If .workflow/current.yml is corrupted:

1. Check for backup: .workflow/current.yml.bak
2. Attempt recovery from manifest:
   • Read project_manifest.json for entity status
   • Scan version-specific tasks directory for task status
   • Reconstruct workflow state
3. If unrecoverable:
   • Report error
   • Suggest starting fresh with /workflow:spawn

ABORT WORKFLOW

If user chooses to abort:

1. Confirm abort: "This will cancel the workflow. Files already created will remain. Continue?"

2. If confirmed:

   • Archive state to .workflow/history/<id>_aborted.yml
   • Clear .workflow/current.yml
   • Report: "Workflow aborted. Created files remain in place."

3. Cleanup options:

   • Offer to rollback created files (if git available)
   • Offer to keep partial implementation