subagent-driven-development

Dispatch isolated subagents for each task with per-task and final reviews.

Updated Jan 12, 2026
One-click install
npx skills add https://github.com/miguelarcjr/shark-ai --skill subagent-driven-development-miguelarcjr
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: subagent-driven-development
Source: https://github.com/miguelarcjr/shark-ai/tree/main/skills/subagent-driven-development
Command: npx skills add https://github.com/miguelarcjr/shark-ai --skill subagent-driven-development-miguelarcjr

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Subagent-driven development solves the challenge of coordinating complex, multi-task work by breaking a plan into independently executed tasks guided by isolated subagents, preserving session context and enabling precise reviews.

Core Features & Use Cases

  • Fresh implementer subagents per task with isolated context to prevent cross-task context bleed.
  • Per-task reviews for specification compliance and code quality, plus a final whole-branch review to consolidate results.
  • Automated task orchestration with global constraints awareness and a ledger-style progress trail to pause and resume sessions.

Quick Start

Dispatch a fresh implementer subagent for each task and perform per-task reviews (spec compliance + code quality) before proceeding to the final whole-branch review.

Frequently Asked Questions about subagent-driven-development

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I coordinate multi-step code tasks without cross-task context bleed?

Subagent-driven development coordinates multi-step code tasks by dispatching fresh implementer subagents per task, ensuring isolated context to prevent cross-task context bleed while tracking progress.

What is the best way to automate code review for specification compliance in a multi-task workflow?

Automated code review for specification compliance is handled by performing per-task diff-based reviews during task orchestration, followed by a final whole-branch consolidation review to verify code quality.

Can I pause and resume a multi-task development session without losing progress?

Yes, you can pause and resume multi-task development sessions because the orchestration maintains a ledger-style progress trail that persists task completion states and global constraints awareness.

How does subagent task dispatch work for independent code execution?

Subagent task dispatch works by allocating an autonomous, independent subagent to execute each specific task from a multi-step plan, applying deterministic orchestration and explicit context separation.

Does subagent-driven development require external dependencies to orchestrate tasks?

No external dependencies are required to orchestrate tasks with subagent-driven development, as it applies in-session to execute multi-step plans using isolated subagents and diff-based reviews.

When should I not use subagent orchestration for software engineering tasks?

You should avoid subagent orchestration for software engineering tasks when a plan cannot be cleanly decomposed into independent tasks, as the workflow relies on explicit context separation and deterministic task dispatch.