Subagent Driven Development

This is a workflow for breaking down implementation plans into independent tasks and dispatching a fresh Claude subagent for each one, with code review between tasks or batches. It works for both sequential work (tasks depend on each other) and parallel execution (investigating multiple unrelated bugs at once). Each subagent gets a clean context, does its work, reports back, then a code reviewer subagent checks the changes before you move on. The interesting part is the quality gate concept: you catch issues immediately after each task instead of discovering a mess at the end. Useful when you have a multi-task plan or 3+ independent failures to investigate, but overkill for simple single-task changes.

skills.sh installs: 1.3K
GitHub stars: 1.8K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add neolabhq/context-engineering-kit --skill subagent-driven-development --agent claude-code

Installs into .claude/skills of the current project.

Files

Subagent-Driven Development

Create and execute plan by dispatching fresh subagent per task or issue, with code and output review after each or batch of tasks.

Core principle: Fresh subagent per task + review between or after tasks = high quality, fast iteration.

Executing Plans through agents:

Same session (no context switch)
Fresh subagent per task (no context pollution)
Code review after each or batch of task (catch issues early)
Faster iteration (no human-in-loop between tasks)

Supported types of execution

Sequential Execution

When you have a tasks or issues that are related to each other, and they need to be executed in order, investigating or modifying them sequentially is the best way to go.

Dispatch one agent per task or issue. Let it work sequentially. Review the output and code after each task or issue.

When to use:

Tasks are tightly coupled
Tasks should be executed in order

Parallel Execution

When you have multiple unrelated tasks or issues (different files, different subsystems, different bugs), investigatin or modifying them sequentially wastes time. Each task or investigation is independent and can happen in parallel.

SKILL.md on GitHub