Is code review mandatory for every task?

Yes, the core principle of this skill relies on a code review after every task to catch issues early and prevent the accumulation of errors.

Which skills are required for this to work?

This skill integrates with writing-plans (to create the roadmap), requesting-code-review (for task validation), and finishing-a-development-branch (for final delivery).

How does this differ from standard plan execution?

Unlike parallel sessions, this stays in the current session but uses fresh subagents for each task to avoid context pollution and enforce quality gates between steps.

What happens if a subagent fails a task?

If a task fails or the reviewer finds issues, a specific fix-subagent is dispatched with detailed instructions to correct the work before moving to the next task.

Subagent-Driven Development

Name: Subagent-Driven Development
Author: franroa

byfranroa

0•

生産性とワークフロー

Orchestrates complex coding plans by dispatching fresh subagents for individual tasks with automated code review gates.

This skill implements a high-integrity workflow for executing multi-task development plans within a single Claude Code session. By isolating each task to a fresh subagent and enforcing a mandatory code review before proceeding, it prevents context pollution and catches architectural errors early. It provides a structured path from plan loading to final verification, ensuring that every code change is tested, reviewed, and aligned with the original project requirements without the need for manual human-in-the-loop oversight between tasks.

主な機能

01Isolated Subagent Execution: Dispatches fresh subagents for every task to maintain clean context and focus.

02Real-time Feedback Loops: Immediately applies review feedback through targeted follow-up subagents.

030 GitHub stars

04Final Architectural Validation: Conducts a comprehensive review of the entire implementation before merging.

05Automated Quality Gates: Integrates mandatory code review cycles after each sub-task implementation.

06Integrated Todo Tracking: Manages task completion status using TodoWrite to maintain continuous progress.

ユースケース

01Implementing test-driven development (TDD) patterns across modular software components.

02Executing complex feature roadmaps that contain multiple independent or semi-independent tasks.

03Maintaining high code quality in long-running sessions without manual context management.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add franroa/chezmoi subagent-driven-development

For use in Claude.ai and ChatGPT

Download Skill