Get the FREE Ultimate OpenClaw Setup Guide β†’

resume-plan

npx machina-cli add skill uukuguy/dev-phase-manager/resume-plan --openclaw
Files (1)
SKILL.md
5.9 KB

Resume Plan

Resume plan execution from saved checkpoint, enabling seamless continuation after /clear.

Use Cases

  • Resume work after clearing context
  • Continue execution from saved progress
  • Switch between different execution modes
  • Recover from interrupted sessions

Execution Steps

1. Load Checkpoint

Read docs/plans/.checkpoint.json:

checkpoint_file="docs/plans/.checkpoint.json"

if [ ! -f "$checkpoint_file" ]; then
  echo "⚠️ Checkpoint file not found"
  echo ""
  echo "Possible reasons:"
  echo "1. /checkpoint-plan was never executed"
  echo "2. Checkpoint has been cleaned up"
  echo ""
  echo "Suggested actions:"
  echo "1. /list-plan - View current plan"
  echo "2. /writing-plans - Create new plan"
  exit 1
fi

2. Load Plan File

Read the plan file path from checkpoint and load plan content into context.

3. Display Progress Summary

Show detailed progress information based on checkpoint content:

πŸ“‹ Resuming Plan Execution

Phase: Phase 5 - MCP Server
Plan: docs/plans/2026-02-22-mcp-server.md
Status: design β†’ execution
Created: 2026-02-22 17:30
Updated: 2026-02-22 17:30

Progress: 0/10 tasks completed
Next: Task 1 - Setup MCP server structure

Notes: Design phase completed, ready for execution

Progress Calculation:

  • If completed_tasks is empty β†’ 0/N tasks completed
  • If tasks are completed β†’ X/N tasks completed (X/N * 100%)
  • Parse total task count N from plan file

4. Select Execution Mode

If execution_mode in checkpoint is null, ask user to choose:

Please select execution mode:

1. subagent-driven-development
   - Execute in current session
   - Spawn new subagent per task
   - Automatic review (spec + code quality)
   - Fast continuous execution
   - Best for: Independent tasks, auto-verifiable

2. executing-plans
   - Execute in current session
   - Batch execution (default 3 tasks)
   - Manual checkpoints
   - Best for: Scenarios requiring manual review

Enter 1 or 2:

If checkpoint already has execution_mode, display and ask whether to continue:

Last execution mode: subagent-driven-development

Continue with same mode? (y/n)
If n, will re-select execution mode

5. Prepare Execution Context

Load the following information into context:

  1. Plan file content: Complete plan document
  2. Progress info: List of completed tasks
  3. Current task: Next task to execute
  4. Execution suggestions: Specific recommendations based on progress

6. Prompt User to Start Execution

Based on user selection, prompt to invoke the appropriate superpowers skill:

βœ… Ready to execute

Plan file loaded into context.

Execute the following command to start:

/subagent-driven-development

The skill will automatically read the plan and start from Task 1.

Or if there are completed tasks:

βœ… Ready to continue execution

Completed: Task 1, Task 2, Task 3
Next: Task 4 - Implement tool handlers

Execute the following command to continue:

/subagent-driven-development

The skill will automatically skip completed tasks and start from Task 4.

Usage Examples

Scenario 1: First execution after design completion

/checkpoint-plan    # Save design phase
/clear
/resume-plan        # Display progress, select execution mode
# Select: 1 (subagent-driven-development)
/subagent-driven-development

Scenario 2: Resume during execution

# After executing 3 tasks
/checkpoint-plan    # Save progress
/clear
/resume-plan        # Display progress: 3/10 completed
# Continue with same mode: y
/subagent-driven-development

Scenario 3: Switch execution mode

/resume-plan
# Last used: subagent-driven-development
# Continue with same mode? n
# Select: 2 (executing-plans)
/executing-plans

Smart Features

1. Auto-detect Plan Changes

If plan file was modified after checkpoint:

⚠️ Plan file has been updated

Checkpoint time: 2026-02-22 17:30
Plan modified: 2026-02-22 18:00

Suggested actions:
1. View plan changes: cat docs/plans/2026-02-22-mcp-server.md
2. Update checkpoint: /checkpoint-plan
3. Continue execution: /resume-plan

Still use old checkpoint? (y/n)

2. Multiple Plan Detection

If multiple plan files exist in docs/plans/:

⚠️ Multiple plan files detected

Checkpoint points to: docs/plans/2026-02-22-mcp-server.md
Other plans:
  - docs/plans/2026-02-21-cognitive-layer.md
  - docs/plans/2026-02-20-skill-vault.md

Use plan from checkpoint? (y/n)
If n, will list all plans for selection

3. Phase Stack Integration

If phase stack file docs/dev/.phase_stack.json exists, display phase info:

πŸ“‹ Resuming Plan Execution

Phase: Phase 5 - MCP Server (active)
Plan: docs/plans/2026-02-22-mcp-server.md
...

⏸️  Suspended phases:
  Phase 4 - Cognitive Layer (60% completed)

Tip: After completing current phase, use /start-phase --resume phase4 to restore

Integration with Third-party Skills

Does not modify third-party skills, integrates through:

  1. Context passing: resume-plan loads plan content into context
  2. Progress hints: Shows completed tasks, hints where to continue
  3. User invocation: Prompts user to manually invoke superpowers skills
resume-plan (user-defined)
  ↓ Load plan into context
  ↓ Display progress summary
  ↓ Prompt user to invoke
  ↓
subagent-driven-development (superpowers)
  or
executing-plans (superpowers)

File Locations

  • Checkpoint file: docs/plans/.checkpoint.json
  • Plan files: docs/plans/YYYY-MM-DD-feature.md
  • Phase stack file: docs/dev/.phase_stack.json (optional)

Notes

  1. No direct invocation: resume-plan doesn't directly call superpowers skills, but prompts user to invoke
  2. Context preparation: Ensure plan content is fully loaded into context
  3. Progress sync: Displayed progress should match actual git history
  4. Flexible selection: Allow users to switch execution modes

Source

git clone https://github.com/uukuguy/dev-phase-manager/blob/main/skills/resume-plan/SKILL.mdView on GitHub

Overview

Resumes a paused planning workflow from a saved checkpoint. It reads the checkpoint, loads the plan, displays progress, and lets you choose an execution mode to continue, enabling seamless recovery after interruptions or clears.

How This Skill Works

The skill loads docs/plans/.checkpoint.json, reads the referenced plan file, and aggregates progress. It then prompts you to select an execution mode (or continues with the existing one), prepares the execution context, and prompts you to start the appropriate execution command to resume from the next task.

When to Use It

  • Resume work after clearing context
  • Continue execution from saved progress
  • Switch between subagent-driven-development and executing-plans modes
  • Recover from interrupted sessions
  • Verify plan consistency after changes and resume

Quick Start

  1. Step 1: /checkpoint-plan to save the current design or progress
  2. Step 2: /clear then /resume-plan to view progress and choose mode
  3. Step 3: Run /subagent-driven-development or /executing-plans to resume

Best Practices

  • Ensure the checkpoint file exists before attempting to resume
  • Review the referenced plan file path and its content loaded into context
  • If execution_mode is null, carefully select a mode that matches your workflow
  • Verify completed_tasks and calculate progress from the plan's total tasks
  • Test resume flow in a safe environment before applying to critical plans

Example Use Cases

  • Scenario 1: First execution after design completion
  • Scenario 2: Resume during execution after saving progress
  • Scenario 3: Switch between subagent-driven-development and executing-plans
  • Scenario 4: Recover from a missing or cleaned checkpoint
  • Scenario 5: Continue after partial progress, starting from the next task

Frequently Asked Questions

Add this skill to your agents
Sponsor this space

Reach thousands of developers β†—