long-task-continuation

Maintain checkpointed state and drift-aware progress for multi-step tasks.

968|43|Updated Apr 30, 2026

Search Tags:#subagents #protocol #resume #evidence #checkpoint #drift #long-task

Ready to run

npx skills add https://github.com/GanyuanRan/Aegis --skill long-task-continuation

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: long-task-continuation
Source: https://github.com/GanyuanRan/Aegis/tree/main/skills/long-task-continuation
Command: npx skills add https://github.com/GanyuanRan/Aegis --skill long-task-continuation

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Long tasks often span multiple steps, contexts, or sessions and risk losing state before completion. This skill provides checkpointing, drift-awareness, and evidence-gated progression to keep work coherent and recoverable.

Core Features & Use Cases

Checkpointed progress across slices and pauses
Drift-aware progression with evidence gating
Structured artifacts and workflow discipline for long-running tasks

Quick Start

State the desired outcome and required baseline refs, then initialize the first checkpoint and baseline usage records before starting the long-task protocol.

Frequently Asked Questions about long-task-continuation

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I resume a long task after a session restart?▼

To resume a long task after a session restart, you use checkpointed state to recover progress. This skill maintains checkpointed state and structured artifacts, allowing multi-step tasks to resume coherently from the last verified boundary without losing prior context.

What is drift-aware progression in long-running tasks?▼

Drift-aware progression in long-running tasks is a mechanism that validates progress against a baseline before continuing. This skill applies drift checks and evidence gating to ensure multi-step work remains coherent and verifiable before advancing to the next checkpoint.

How do I manage state when handing off subagents in a multi-step protocol?▼

To manage state when handing off to subagents, you establish clear boundaries and checkpointed progress. This skill provides a long-task protocol that maintains structured artifacts and verifiable evidence, ensuring seamless state handoff across different architectures.

When do I need evidence gating for checkpointed workflows?▼

You need evidence gating for checkpointed workflows when tasks span multiple sessions and require verifiable progress before completion. This skill enforces evidence-gated progression to ensure that each checkpoint is backed by a baseline usage record and validated artifacts.

Can I use this checkpointing protocol for tasks spanning multiple architectures?▼

Yes, you can use this checkpointing protocol for tasks spanning multiple architectures. The skill is explicitly designed for multi-step tasks that may restart or hand off to subagents across different architectures, maintaining coherent state through structured artifacts.

What are the limitations of using evidence-gated progression for task completion?▼

A key limitation of this evidence-gated progression is that it controls completion without executing plans or dispatching subagents itself. The skill provides drift checks and checkpointed state management, but requires external execution systems to carry out the actual task steps.