What is 'compaction' and why do I need this skill for it?

Compaction occurs when Claude summarizes long histories to stay within context limits. While history is summarized, this skill ensures your task list, dependencies, and agent assignments remain intact.

Why should I prefix tasks with [agent] tags?

Explicitly naming the agent (e.g., [engineer] or [tester]) in the task subject ensures that even after the conversation is summarized, the specific responsibility for each task is recoverable.

How do task dependencies work in this skill?

You can use 'blockedBy' and 'blocks' fields to encode logical sequences, allowing the AI to automatically identify which tasks are actionable and which must wait for previous steps.

Does this replace my issue tracker like Linear or GitHub?

No, this is for local session state and working memory. You should pull work from your tracker into tasks for execution and sync results back once completion boundaries are reached.

Session Task Manager

Name: Session Task Manager
Author: outfitter-dev

byoutfitter-dev

•

生产力与工作流

Maintains session state and task coordination across context compactions to ensure complex workflows remain focused and organized.

This skill provides a structured task management system designed to survive context compaction during long AI coding sessions. It enables users and subagents to track local execution state, preserve intent through explicit agent assignments, and manage complex multi-step dependencies. By maintaining a persistent record of what is done, what is in progress, and who is responsible, it ensures that Claude never loses track of the project's goal even as the conversation history is summarized and older messages are removed from the active context window.

主要功能

01Real-time status tracking (pending, in_progress, completed)

02Explicit subagent assignment via subject prefixes

03Multi-step workflow sequencing with blocked/blocking dependencies

04Compaction-resistant task persistence

05Intent recovery and execution context preservation

062 GitHub stars

使用场景

01Managing parallel workstreams between different subagents like engineers and reviewers

02Preserving complex project state during long-running sessions that exceed context limits

03Coordinating multi-step migrations where context loss would cause repetition

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add outfitter-dev/outfitter maintain-tasks

For use in Claude.ai and ChatGPT

Download Skill