subagent-driven-development

Executes implementation plans by dispatching fresh subagents per task with two-stage review.

1|Updated Dec 11, 2025
One-click install
npx skills add https://github.com/HarmAalbers/claude-requirements-framework --skill subagent-driven-development-harmaalbers
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: subagent-driven-development
Source: https://github.com/HarmAalbers/claude-requirements-framework/tree/main/plugins/requirements-framework/skills/subagent-driven-development
Command: npx skills add https://github.com/HarmAalbers/claude-requirements-framework --skill subagent-driven-development-harmaalbers

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Executing a multi-task implementation plan in a single session often leads to context pollution, skipped reviews, and unverified claims of completion. This Skill orchestrates plan execution by dispatching a fresh subagent per task and enforcing spec compliance and code quality reviews after each task. ## Core Features & Use Cases - Fresh subagent per task: Each task runs in a clean context with a detailed brief containing verified codebase facts, preventing confusion and silent guessing. - Two-stage review loop: A spec compliance reviewer confirms the code matches requirements before a code quality reviewer assesses correctness, testing, and maintainability. - Diff-grounded verification: Reviewers re-run tests, inspect the full git diff-stat, and verify commits exist rather than trusting implementer reports. - Use Case: Given a written implementation plan with five independent tasks, dispatch implementer subagents sequentially, review each task's diff, and finish with a final whole-implementation code review before merging the branch. ## Quick Start Ask the AI to execute the implementation plan using subagent-driven development with a fresh implementer subagent and two-stage review for each task.

Frequently Asked Questions about subagent-driven-development

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

FAQPage Schema
How do I execute an implementation plan with subagents?▼

Extract all tasks from the plan, then dispatch one implementer subagent per task with a brief containing the full task text and verified codebase facts. After each task, run a spec compliance review followed by a code quality review before moving to the next task.

What is the difference between subagent-driven development and executing plans?▼

Subagent-driven development runs in the same session with a fresh subagent per task and automatic review checkpoints. Executing plans uses a parallel session with a handoff, which suits workflows needing session isolation or human-in-the-loop between tasks.

Why should reviewers read the diff instead of the implementer's report?▼

Reports routinely claim all checks pass while the diff contains bugs, formatter churn, or committed build artifacts. Reviewers must re-run tests, inspect the full git diff-stat, check git status, and verify the commit exists before approving.

Can I dispatch multiple implementer subagents in parallel?▼

No, parallel implementation subagents cause conflicts and are explicitly prohibited. Tasks are executed sequentially, with each task completing both review stages before the next implementer is dispatched.

When should the orchestrator do the review itself instead of using reviewer subagents?▼

The orchestrator may perform both review stages when it already holds the plan and seams context and the diff is small enough to read inline. If the diff becomes too large to read fully, fall back to fresh reviewer subagents.