skills.rest
pgoell
pgoellCOMMUNITY
Explore all skills by this author →
Website𝕏@pagoellGitHub

subagent-driven-development

Executes implementation plans by dispatching bounded tasks to subagents with spec and quality review gates.

Updated Mar 18, 2026
Search Tags:#tdd#code review#parallel agents#workflow orchestration#subagents#implementation plan#task dispatch
One-click install
npx skills add https://github.com/pgoell/pgoell-claude-tools --skill subagent-driven-development-pgoell
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: subagent-driven-development
Source: https://github.com/pgoell/pgoell-claude-tools/tree/main/plugins/workbench/skills/subagent-driven-development
Command: npx skills add https://github.com/pgoell/pgoell-claude-tools --skill subagent-driven-development-pgoell

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Executing a multi-slice implementation plan in one session often leads to context bloat, scope creep, and unreviewed changes. This Skill coordinates fresh subagents per task slice with enforced review gates so each change is verified before moving on. ## Core Features & Use Cases - Slice-Based Task Dispatch: Assigns one implementation agent per plan slice with explicit file ownership, acceptance criteria, and verification commands. - Two Review Gates: Runs a spec compliance reviewer followed by a code quality reviewer before any slice is marked complete. - Parallel Dispatch Rules: Defines when tasks can run in parallel (disjoint write scopes, independent domains) and when they must stay sequential. - Use Case: You have a plan with five slices from a planning session. This Skill walks each slice through an implementation agent, both review gates, deviation tracking, and a final full verification pass before completion. ## Quick Start Execute the implementation plan using subagent-driven development, dispatching one agent per slice with review gates.

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?▼

Assign one implementation agent per plan slice with explicit file ownership, acceptance criteria, and verification commands. After each agent returns, run a spec compliance review then a code quality review before marking the slice complete.

When should I run coding subagents in parallel?▼

Dispatch agents in parallel only when tasks cover independent domains, have disjoint write scopes or are read-only, and no task depends on another's result. Avoid parallelism when tasks edit the same files, schemas, or lockfiles.

What review steps should follow subagent code changes?▼

Run two sequential gates: a spec compliance reviewer confirming the implementation matches the plan, then a code quality reviewer checking maintainability, tests, and integration risk. Never start quality review before spec compliance passes.

What should I do when a subagent returns BLOCKED status?▼

Change something before retrying: provide missing context, split the task into smaller pieces, use a more capable model, or stop if the plan itself is wrong. Do not resubmit the identical prompt unchanged.

Does this workflow work in both Claude Code and Codex?▼

Yes. Claude Code uses the Agent tool for implementation and review agents, while Codex uses spawn_agent for sidecar tasks. For urgent blocking work in Codex, implement locally while preserving the same task loop discipline.

Related Skills

Explore More Software Engineering Skills→
johangm90johangm90

spex-qa

Perform QA, security reviews, and code reviews for AI agent framework development.

Community
Advanced
mykaarmamykaarma

verification-before-completion

Enforces verification-before-completion workflow for accurate project statuses.

Official
Intermediate
tim-hubtim-hub

test-driven-development

Enforce test-first development with failing tests before implementation.

Community
Intermediate
m16khbm16khb

subagent-driven-development

Dispatch fresh subagents per task with code review and test gates.

Community
Advanced