Can I run more than two subagents at once?

No, the skill is intentionally capped at two parallel subagents to maintain high quality and ensure that the code review process remains manageable and effective.

How does this differ from standard Subagent-Driven Development?

While standard development is sequential, this skill runs up to two subagents in parallel for independent tasks, cutting down wall-clock time by up to 40%.

What happens if a code review identifies issues?

The workflow pauses to dispatch follow-up subagents specifically to fix the issues before the manifest is updated and the next batch of tasks begins.

Why does the skill use a 'fresh' subagent for each task?

Using a fresh subagent prevents context pollution, ensuring the AI stays focused on the specific task requirements without being distracted by unrelated implementation details.

Is a manifest file required to use this skill?

Yes, you must run the 'decomposing-plans' skill first to generate the manifest.json and individual task files that this skill uses for execution.

Parallel Subagent Developer

Name: Parallel Subagent Developer
Author: seanGSISG

byseanGSISG

0•

生産性とワークフロー

Accelerates development by dispatching parallel subagents for independent tasks with integrated code review quality gates.

This skill optimizes the AI development workflow by executing complex, decomposed plans through parallelized subagents. By splitting independent tasks into batches of two, it significantly reduces total implementation time while maintaining high code quality through mandatory review checkpoints between every batch. It utilizes a manifest-driven architecture to provide fresh context for every task, preventing context pollution and ensuring strict adherence to Test-Driven Development (TDD) principles and project-specific best practices.

主な機能

01Automated feedback loops for fixing review-identified issues

02Manifest-based task tracking and status management via JSON

030 GitHub stars

04Isolated subagent contexts to prevent cross-task context pollution

05Parallel execution of up to two subagents for independent tasks

06Mandatory code review quality gates between implementation batches

ユースケース

01Scaling development speed on large features with independent sub-modules

02Implementing complex plans where task isolation is critical for code quality

03Automating the TDD lifecycle across multiple concurrent workstreams

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add seangsisg/crispy-claude parallel-subagent-driven-development

For use in Claude.ai and ChatGPT

Download Skill