How does this differ from standard plan execution?

Unlike basic execution, this skill uses fresh subagents for every task to ensure zero context leakage and enforces a strict two-stage review process before any task is marked complete.

What happens if a subagent gets stuck?

The skill handles statuses like BLOCKED or NEEDS_CONTEXT, allowing the controller to provide missing info, break down the task, or escalate to a more powerful model.

Does this skill require specific project setups?

It is most effective when used with a structured implementation plan and ideally within a git worktree to maintain an isolated and clean workspace.

Does it work with different LLM models?

Yes, the skill encourages using faster, cheaper models for mechanical tasks and more capable models for complex architecture and review tasks to optimize performance.

Subagent-Driven Development

Name: Subagent-Driven Development
Author: saar120

bysaar120

0•

Productivity & Workflow

Executes complex implementation plans by dispatching specialized subagents with rigorous two-stage review cycles for every task.

Subagent-Driven Development transforms how Claude handles multi-task implementation plans by delegating individual tasks to fresh, isolated subagents. This approach prevents context pollution and ensures high-quality output through a mandatory two-stage review process: first verifying spec compliance, then evaluating code quality. It is ideal for executing well-defined, independent tasks within a single session, combining the speed of AI automation with the reliability of structured quality gates and Test-Driven Development (TDD) principles.

Key Features

01Intelligent model selection based on task complexity for cost optimization

02Automated task extraction and Todo management for implementation plans

030 GitHub stars

04Two-stage review process verifying spec compliance and code quality

05Integrated support for Test-Driven Development (TDD) via subagent roles

06Fresh context isolation for every task to prevent session pollution

Use Cases

01Refactoring large codebases where tasks are independent and require strict spec adherence

02Executing multi-step feature implementations from a pre-defined architectural plan

03Automating high-quality code production in fast-paced development cycles with minimal human-in-the-loop

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add saar120/skillgate-registry subagent-driven-development

For use in Claude.ai and ChatGPT

Key Features

01Intelligent model selection based on task complexity for cost optimization

02Automated task extraction and Todo management for implementation plans

030 GitHub stars

04Two-stage review process verifying spec compliance and code quality

05Integrated support for Test-Driven Development (TDD) via subagent roles

06Fresh context isolation for every task to prevent session pollution

Use Cases

01Refactoring large codebases where tasks are independent and require strict spec adherence

02Executing multi-step feature implementations from a pre-defined architectural plan

03Automating high-quality code production in fast-paced development cycles with minimal human-in-the-loop

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add saar120/skillgate-registry subagent-driven-development

For use in Claude.ai and ChatGPT