How do I assign tasks to different agents?

Use a bracketed prefix in the task subject, such as [engineer], [reviewer], or [tester]. This explicit assignment ensures that the specific role responsible for the task is clearly identifiable even after the conversation history is condensed.

Does this replace project management tools like Linear or GitHub?

No. This skill is for local execution state—the immediate 'what am I doing now' and 'what's next.' You should still pull work from your primary issue tracker at the start and sync your completed work back to it at the end.

When should I use the maintain-tasks skill?

You should use it for any non-trivial work involving three or more steps, when coordinating multiple subagents, or when performing complex work that requires sequencing and dependency tracking.

How does this skill survive context compaction?

While standard conversation history is summarized during compaction, tasks created with this skill remain intact with full descriptions, statuses, and dependencies, acting as a permanent working memory for the session.

Agent Task Management

Name: Agent Task Management
Author: outfitter-dev

byoutfitter-dev

•

生産性とワークフロー

Maintains session continuity and agent coordination by tracking tasks and dependencies through context compaction.

概要

The maintain-tasks skill provides a robust framework for managing complex, multi-step workflows within Claude Code, ensuring that intent and progress are preserved even when conversation history is summarized. By implementing explicit subagent assignments and task dependencies, it allows projects to be sequenced effectively across different roles like engineering, reviewing, and testing. It acts as a persistent 'working memory' that survives the natural lifecycle of an AI session, preventing the loss of nuanced instructions and state during long-running development tasks.

主な機能

Persistent task tracking that survives context compaction
Explicit subagent assignment via bracketed prefixes (e.g., [engineer])
Local execution state management for session-specific work
18 GitHub stars
Dependency mapping with 'blockedBy' and 'blocks' relationships
Standardized sync patterns for Linear and GitHub workflows

ユースケース

Managing complex refactoring projects that exceed standard context limits
Preserving specific execution intent and 'done criteria' across long sessions
Coordinating sequential handoffs between research, engineering, and QA agents

概要

主な機能

Persistent task tracking that survives context compaction
Explicit subagent assignment via bracketed prefixes (e.g., [engineer])
Local execution state management for session-specific work
18 GitHub stars
Dependency mapping with 'blockedBy' and 'blocks' relationships
Standardized sync patterns for Linear and GitHub workflows

ユースケース

Managing complex refactoring projects that exceed standard context limits
Preserving specific execution intent and 'done criteria' across long sessions
Coordinating sequential handoffs between research, engineering, and QA agents