development-task-routing

Plans development task decomposition, executor assignment, and integration acceptance for multi-agent workflows.

Updated Mar 17, 2026
One-click install
npx skills add https://github.com/DJackyB/Project_Rent --skill development-task-routing-djackyb
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: development-task-routing
Source: https://github.com/DJackyB/Project_Rent/tree/main/.claude/skills/development-task-routing
Command: npx skills add https://github.com/DJackyB/Project_Rent --skill development-task-routing-djackyb

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Starting a new development task without a clear plan leads to rework, unclear ownership, and integration failures. This Skill structures the kickoff phase by judging task scale, splitting work into assignable subtasks, dispatching them to the right executors, and defining integration and acceptance checkpoints. ## Core Features & Use Cases - Plan-mode judgment and task sizing: Automatically decides whether a task needs plan mode and whether it is small, medium, or large before any implementation begins. - Executor-based dispatching: Assigns subtasks to the main AI, Claude Code, Codex sub-agents, Antigravity, or humans based on task nature rather than even workload distribution. - Integration and acceptance orchestration: Enforces explicit task packages, handoff criteria, integration checkpoints, and final acceptance so subtasks do not silently fail to connect. - Use Case: When starting a new feature that touches core logic, UI configuration, and scene wiring, use this Skill to split the work into phases, assign core logic to Claude Code, parallel slices to Codex sub-agents, manual wiring to Antigravity, and reserve integration and acceptance for the main AI. ## Quick Start Ask the AI to use development-task-routing to judge whether your new task needs plan mode, split it into subtasks, and assign each part to the right executor.

Frequently Asked Questions about development-task-routing

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

FAQPage Schema
How do I split a development task across multiple AI agents?▼

Split by phase first, then by subtask, and assign each subtask to the executor best suited to its nature: core logic to Claude Code, independent slices to Codex sub-agents, manual wiring to Antigravity or humans. Each subtask needs a task package with goals, boundaries, and acceptance criteria.

When should a coding task enter plan mode before execution?▼

Enter plan mode when the task has multiple dependent phases, competing approaches with non-obvious consequences, multiple executors, or high rework cost. Single-point bugs and small well-scoped fixes can be executed directly without planning overhead.

Which tasks should go to Claude Code versus Codex sub-agents?▼

Claude Code handles core business logic, critical data flows, and high-risk refactoring where quality matters most. Codex sub-agents handle independent, low-coupling slices, read-only analysis, test additions, and parallel sidecar work that does not block key decisions.

Should small bug fixes be split into subtasks?▼

No. Single-point bugs, small scoped adjustments, and local test additions should be done directly. Splitting is only required for multi-module work, mixed logic and UI changes, or tasks needing multiple executors, where coordination cost justifies the structure.

Why avoid fallback logic and default values in the main workflow?▼

Silent fallbacks and default-value patching hide real problems like missing configuration or broken contracts until later stages. The Skill requires errors to surface early at startup, integration, or acceptance, allowing controlled degradation only for explicitly optional capabilities.

What decisions must be confirmed with the user before execution?▼

Large decisions affecting scope, cost, architecture, module boundaries, long-term automation, or maintenance approach must be asked first. Only local, reversible, low-risk implementation details that do not change the overall direction can be decided by default.