Subagent-Driven Development FAQs

Question 1

What happens if a subagent fails a task review?

Accepted Answer

The system triggers a fix-loop where the implementer subagent addresses the specific reviewer feedback and submits the work for re-review until all quality gates are passed.

Question 2

How does this differ from standard plan execution?

Accepted Answer

Unlike standard execution, this skill stays within the same session but uses fresh subagents for each task to prevent context pollution while automating a mandatory two-stage review process.

Question 3

Why is the review split into two stages?

Accepted Answer

The spec compliance review ensures the code actually does what was requested, while the code quality review focuses on best practices and maintainability, preventing logic errors from being buried in stylistic fixes.

Question 4

Can I use this for parallel development?

Accepted Answer

No, this skill is designed for sequential task execution within a single session to avoid git conflicts and ensure logical progression through the implementation plan.

Question 5

Do I need to manually review every task?

Accepted Answer

The skill automates subagent-led reviews, but you remain in the loop to answer clarifying questions or provide context if an implementer subagent gets stuck.

Subagent-Driven Development

主要功能

使用场景

Subagent-Driven Development

主要功能

使用场景