mcpField
L
long-running-task-management

long-running-task-management

github.com/archieindianAgent workflows7367 installsUpdated 26w ago
Repository

About long-running-task-management

A skill for agents managing lengthy tasks, including work that spans sessions or many files. It records stages and checkpoints in a state file so work can resume after interruption.

Key features

  1. 01Create a task state file
  2. 02Record stage status and checkpoints
  3. 03Resume at pending stages after interruption
  4. 04Run full verification at completion

Use cases

  1. 01Track tasks estimated to take over 30 minutes
  2. 02Continue work across multiple sessions
  3. 03Coordinate changes across many directories

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/archieindian/openclaw-superpowers --skill long-running-task-management

After installing, type/long-running-task-managementin Claude Code to confirm it loaded.

Bundled filesSTATE_SCHEMA.yaml

Details

From the project README

Long-Running Task Management

State file: ~/.openclaw/skill-state/long-running-task-management/state.yaml

When to Use

  • Task estimated at more than 30 minutes
  • Task will span multiple sessions
  • Task modifies many files across multiple directories

Starting a Task

  1. Write initial state to the state file:
    • task_id: short kebab-case name
    • status: in_progress
    • description: one-sentence goal
    • stages: ordered list with status: pending for each
    • started_at: current timestamp
  2. Begin the first stage

At Each Checkpoint

  1. Complete the stage
  2. Run tests/verification
  3. Update state file: mark stage status: complete, write checkpoint (what's stable now), write next_action (first thing to do on resume), update last_updated
  4. Commit progress to git if applicable

Resume After Interruption

  1. Read the state file
  2. Check status and next_action
  3. Continue from the next pending stage — do NOT start over

Completion

  1. Update state: status: complete, final checkpoint
  2. Run full verification

Cron Wakeup Behavior

On each 15-minute wakeup:

  • Read state file
  • If status: in_progress and last_updated is stale (>30 min ago): log a checkpoint update to daily memory
  • If status: complete or no active task: skip

Comments

Markdown supported