sdd

Dispatch fresh subagents for each task with isolated context and two-stage reviews.

15|1|Updated Mar 29, 2026
One-click install
npx skills add https://github.com/duronext/jig --skill sdd-duronext
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: sdd
Source: https://github.com/duronext/jig/tree/main/core/skills/sdd
Command: npx skills add https://github.com/duronext/jig --skill sdd-duronext

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Orchestrating complex implementation plans can overwhelm a single agent. This Skill delegates each task to a fresh subagent, ensuring isolation of context and avoiding cross-task leakage.

Core Features & Use Cases

  • Per-task subagent dispatch with dedicated execution context for each task.
  • Two-stage reviews per task: spec compliance followed by code quality.
  • Flexible execution: serial processing for coupled tasks or when agent teams are unavailable.

Quick Start

Dispatch a fresh subagent for each task and ensure it passes spec compliance before code quality review.

Frequently Asked Questions about sdd

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

FAQPage Schema
How do I execute multi-task implementation plans without overwhelming a single agent?

You can dispatch a fresh subagent for each task to execute plan steps with isolated context, preventing cross-task leakage. This approach coordinates per-task architecture context and dependencies, applying a two-stage review before completion to manage complex plans effectively.

What is a two-stage review for code quality and spec compliance?

A two-stage review enforces spec compliance followed by code quality before completing a dispatched task. This mechanism ensures the subagent's output meets the implementation requirements first, then validates the code quality standards before finalizing the execution.

How do I prevent context leakage between coupled tasks during subagent dispatch?

To prevent context leakage, dispatch a fresh subagent for each task to ensure isolated execution context. For coupled tasks, use serial processing to maintain order and isolation, ensuring per-task architecture context and dependencies are managed without cross-task interference.

When should I use serial processing instead of agent teams for task dispatch?

Use serial processing for task dispatch when handling coupled tasks that require strict sequential execution or when agent teams are unavailable. This ensures per-task dependencies are respected and maintains isolated context for each step of the implementation plan.

Can I coordinate subagents for multi-task plans requiring different execution contexts?

Yes, you can coordinate fresh subagents for multi-task plans by dispatching each task with dedicated execution context. This supports both serial processing for coupled tasks and agent teams, enforcing two-stage reviews to ensure spec compliance and code quality before completion.

Why does my implementation plan fail code quality checks after subagent execution?

Implementation plans fail code quality checks if the two-stage review is not properly enforced. You must ensure each dispatched subagent passes spec compliance first, then undergoes code quality review before the task is marked complete, maintaining isolated context throughout.