skills.rest
treejh
treejhCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

startproject

Coordinate multi-agent research, planning, implementation, and review for new software projects.

Updated Apr 9, 2026
Search Tags:#codex#claude#multi-agent#orchestration#code-review#task-creation#project-kickoff
One-click install
npx skills add https://github.com/treejh/ai-agent-skills --skill startproject-treejh
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: startproject
Source: https://github.com/treejh/ai-agent-skills/tree/main/skills/claude-orchestrator/.claude/skills/startproject
Command: npx skills add https://github.com/treejh/ai-agent-skills --skill startproject-treejh

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill addresses the friction of starting a new repository feature by coordinating research, planning, implementation tasking, and multi-session review across multiple specialist agents to preserve context and improve quality.

Core Features & Use Cases

  • Automated repository research: Run a background Gemini analysis to surface architecture, relevant files, and technical considerations.
  • Requirements gathering and planning: Interactive Claude-driven prompts collect user goals and draft an implementation plan.
  • Design and post-implementation review: Use Codex via subagents for design critique and diff-based multi-session code review.
  • Task creation and persistence: Generate actionable task lists and update CLAUDE.md to persist project context across sessions.
  • Use Case: Kick off a non-trivial feature by producing a vetted plan, task breakdown, and scheduling independent reviews to reduce bias and catch issues early.

Quick Start

Start a new feature called "user-auth" by analyzing the repository for relevant files, drafting requirements and an implementation plan, creating a task list, and scheduling a multi-session review with Codex.

Frequently Asked Questions about startproject

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

FAQPage Schema
How do I kick off a new software project using multi-agent orchestration?▼

Multi-agent orchestration kicks off a project by using Gemini for repository analysis, Claude for requirements gathering and task creation, and Codex for design and code review. This coordinates research, planning, implementation, and review across multiple specialist agents.

What is the best way to automate repository research before starting a feature?▼

Automated repository research is performed by running a background Gemini analysis to surface architecture, relevant files, and technical considerations. This provides the necessary context before Claude gathers requirements and drafts an implementation plan.

Can I use Claude and Codex for diff-based multi-session code review?▼

Yes, Codex via subagents performs diff-based multi-session code review for post-implementation review. Claude handles task creation and persistence, updating CLAUDE.md to persist project context across these independent review sessions.

Does multi-agent project kickoff work for repository-based feature development?▼

Yes, this multi-agent project kickoff applies specifically to repository-based feature development. It reduces bias and catches issues early by scheduling independent design critiques and reviews using coordinated subagents.

How do I persist project context across multiple agent sessions?▼

To persist project context across multiple agent sessions, the workflow generates actionable task lists and updates CLAUDE.md. Automated subagent research outputs are also saved directly to .claude/docs to maintain context.

Related Skills

Explore More Software Engineering Skills→
Minara-AIMinara-AI
6

media-setup

Configure platform connections, API keys, and environment dependencies for media-agent installations.

Official
Advanced
refactory-langrefactory-lang

speckit-review

Review code against specifications and update task statuses in tasks.md.

Official
Advanced
RoyMcCrainRoyMcCrain
1

codex

Run Codex CLI and Claude's code-reviewer subagent for parallel code review.

Community
Intermediate
SteelMorganSteelMorgan
87

task-breakdown-linear

Generate a JSON task breakdown with dependencies and self-checks for linear execution.

Community
Intermediate