task-dispatch

Splits coding tasks into independent subtasks executed in parallel by sub-agents and merges results.

2|Updated Jun 8, 2026
One-click install
npx skills add https://github.com/HACK-WU/skills --skill task-dispatch-hack-wu
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: task-dispatch
Source: https://github.com/HACK-WU/skills/tree/main/skills/task-dispatch
Command: npx skills add https://github.com/HACK-WU/skills --skill task-dispatch-hack-wu

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Completing multi-file or multi-module coding tasks with a single agent is slow and error-prone over long contexts. This Skill decomposes such tasks into independent subtasks, runs them in parallel with sub-agents, and merges the outputs with explicit conflict handling and consistency checks. ## Core Features & Use Cases - Parallel Subtask Decomposition: Splits a task by file/module, independent operations, vertical slices, or responsibility layers, then validates independence across file, interface, shared-resource, and implicit-import dimensions. - Batch Scheduling & Isolated Execution: Groups subtasks into dependency-ordered batches, runs sub-agents that write code to isolated directories, and uses contract files for weak dependencies. - Merge & Consistency Verification: The main agent merges outputs into the project source tree, handles file/interface/logic conflicts in tiers, and verifies compilation, imports, interface consistency, and acceptance criteria. - Use Case: You need to implement user, order, and payment modules at once. The Skill splits them into three fully independent subtasks, launches three sub-agents in parallel, merges their code, and validates the integrated result. ## Quick Start Ask the agent to split the task of implementing the user, order, and payment modules into parallel subtasks and execute them with sub-agents using task-dispatch.

Frequently Asked Questions about task-dispatch

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

FAQPage Schema
How do I run multiple coding subtasks in parallel with AI agents?

Split the task into independent subtasks, group them into dependency-ordered batches, and launch one sub-agent per subtask in the same batch. Each sub-agent writes code to an isolated directory, and the main agent merges the outputs afterward.

When should a coding task not be split for parallel execution?

Avoid splitting when changes are confined to a single file, steps form a strict serial dependency chain, the work needs a global architectural view, or subtasks write to the same core data structure. In those cases a single agent executes directly via the fast path.

How are file conflicts between parallel sub-agents resolved?

The main agent detects when multiple subtasks produce the same file and performs a three-way merge against the original source. Overlapping edits that cannot be merged semantically are flagged for user decision rather than silently overwritten.

How do sub-agents handle interface dependencies between subtasks?

Weak dependencies are handled contract-first: the main agent writes interface contract files before launching sub-agents, and sub-agents must follow them strictly. Sub-agents can also confirm signatures via messages without exchanging large code blocks.

What is the difference between task-dispatch and work-breakdown?

work-breakdown only produces a work-item list without executing anything, while task-dispatch splits the task, executes subtasks in parallel with sub-agents, merges the code, and delivers an integrated result.

What happens if a sub-agent fails or times out during parallel execution?

A timed-out sub-agent's partial output is excluded from merging and marked as unfinished, while other subtasks continue unaffected. The main agent then retries the subtask or completes it manually.