subagent-driven-development

Dispatch subagents per task with dependency-aware waves and self-review.

9|2|Updated Mar 12, 2026

Search Tags:#workflow #orchestration #subagent #parallel-execution #audit-trail #self-review #deterministic-development

Ready to run

npx skills add https://github.com/iclinic/afyapowers --skill subagent-driven-development-iclinic

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: subagent-driven-development
Source: https://github.com/iclinic/afyapowers/tree/main/src/skills/subagent-driven-development
Command: npx skills add https://github.com/iclinic/afyapowers --skill subagent-driven-development-iclinic

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Subagent-Driven Development helps teams execute implementation plans by dispatching independent subagents per task within the current session, enabling parallel execution, fast iteration, and built-in self-review with concerns collection for later audit.

Core Features & Use Cases

Fresh subagent per task with self-review and concerns collection to accelerate iteration.
Wave-based parallel execution with dependency awareness and file-overlap checks.
Suitable for deterministic development workflows that require auditing and traceability.

Quick Start

Instantiate a new feature plan and let subagents execute tasks in parallel, reporting their status and concerns.

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 parallel subagents?▼

Executing an implementation plan with parallel subagents involves dispatching a fresh subagent per task using dependency-aware waves. The system computes ready-sets and validates file-overlap to ensure safe parallel execution, collecting self-review concerns for later audit.

What is dependency-aware wave-based execution for development tasks?▼

Dependency-aware wave-based execution groups independent development tasks to run in parallel, computing ready-sets and applying cycle checks per wave. Tasks with dependencies wait, ensuring safe parallel execution while accelerating iteration across the implementation plan.

How do subagents handle self-review and concerns collection during development?▼

Subagents handle self-review by evaluating their own task implementation before returning a structured status summary. They collect any concerns encountered during execution, logging them into an audit trail for later review and deterministic development traceability.

How does file-overlap validation prevent conflicts during parallel task execution?▼

File-overlap validation prevents conflicts during parallel task execution by checking if independent subagents target the same files. It computes ready-sets per wave and blocks overlapping tasks, ensuring safe parallel execution without write conflicts.

What do the structured status summaries DONE, DONE_WITH_CONCERNS, NEEDS_CONTEXT, and BLOCKED mean?▼

Structured status summaries report task completion states: DONE indicates success, DONE_WITH_CONCERNS flags issues for audit, NEEDS_CONTEXT requests missing information, and BLOCKED shows unresolvable dependencies, enabling deterministic development tracking.

When should I not use subagent-driven parallel execution for implementation plans?▼

You should not use subagent-driven parallel execution when tasks have heavy file-overlap or circular dependencies that prevent wave-based separation. If cycle checks detect unresolvable blockers, sequential execution is required to ensure safe development.