Subagent-Driven Development FAQs

Question 1

Does this work with Test-Driven Development?

Accepted Answer

Absolutely. Subagents are instructed to follow TDD principles for each task, ensuring that implementations are verified by tests before they reach the review stage.

Question 2

Can I use different AI models for different tasks?

Accepted Answer

Yes, the skill is designed to use cheaper, faster models for mechanical tasks and more powerful models for complex architectural or integration tasks to save on costs.

Question 3

How does this differ from standard plan execution?

Accepted Answer

Unlike standard execution which might handle tasks sequentially in one thread, this skill dispatches fresh subagents per task to prevent context pollution and uses a mandatory two-stage review gate for every change.

Question 4

What happens if a subagent gets stuck or blocked?

Accepted Answer

The workflow includes explicit handling for 'BLOCKED' or 'NEEDS_CONTEXT' statuses, allowing the user to provide information or upgrade the model reasoning without losing progress.

Question 5

Why is there a two-stage review process?

Accepted Answer

The first stage ensures the code actually does what the spec requested (Spec Compliance), while the second stage ensures the code is maintainable and follows best practices (Code Quality).

Subagent-Driven Development

主要功能

使用场景

Subagent-Driven Development

主要功能

使用场景