Capafy
Capafy Capafy
AI Development Director

AI Development Director

@Yuu Kamegai
Turn product ideas into implementation-ready plans before your AI agent writes a single line of code. AI Development Director produces structured requirements, architecture decisions, database schemas, implementation tasks, coding-agent prompts, review checklists, and launch checklists. This is not a code-generation prompt. It is the planning system that runs before code generation.
Rating
No Ratings
Sold
0
How to use
Download

AI Development Director

A reusable skill for Claude Code, OpenAI Codex, Cursor, Capafy, and other AI coding agents.

Transforms vague product ideas into structured, implementation-ready development plans. Acts as an AI technical product manager — not a code generator.

The Problem This Solves

AI coding agents are good at writing code. They are bad at deciding what to build, in what order, and with what architecture.

Without planning, agents tend to:

  • Start implementing before requirements are clear
  • Make arbitrary architecture decisions mid-project
  • Produce inconsistent database schemas
  • Miss edge cases, auth flows, and payment failure paths
  • Generate code that works for the happy path and breaks everywhere else

This skill forces a planning step before any implementation begins.

What It Produces

For any product idea, the skill generates:

  1. Structured requirements — target users, core workflow, monetization, auth, scale
  2. Constraint analysis — team, timeline, compliance, operational limits
  3. User flow definitions — every actor, every path, every edge case
  4. Architecture proposal — stack selection with explicit justification
  5. Database schema — tables, columns, relationships, indexes, RLS policies
  6. Task breakdown — sequential, agent-executable implementation tasks
  7. Risk analysis — technical, product, cost, and security risks with mitigations
  8. Coding agent prompts — one implementation-specific prompt per task
  9. Review checklist — project-specific, severity-rated
  10. Launch checklist — phase-gated go/no-go criteria

Default Stack

The skill recommends this stack for most web MVPs:

Layer Technology
Framework Next.js App Router + TypeScript
Database Supabase (PostgreSQL)
Authentication Supabase Auth or Clerk
Payments Stripe
File Storage Supabase Storage
Deployment Vercel
Analytics PostHog

Deviations are explicitly justified in the architecture output.

Installation

Capafy

Upload SKILL.md as a new skill in your Capafy workspace. Include the resources/ directory as skill attachments for full reference support.

Claude Code

Copy SKILL.md to your Claude Code project skills directory:

.claude/skills/ai-development-director.md

Or add as a project-level instruction in CLAUDE.md:

When planning new features or apps, follow the workflow in skills/ai-development-director.md

OpenAI Codex / ChatGPT

Paste the contents of SKILL.md as a system prompt or custom instruction. Attach resource files as reference documents.

Cursor

Add SKILL.md to your .cursor/rules/ directory. Reference resource files in the rule body.

Usage

Basic trigger

Describe a product idea to an agent running this skill:

I want to build a subscription-based tool that lets marketing teams
generate SEO blog posts from their existing content using AI.

The agent will work through all 10 planning steps before producing any code.

Structured trigger

Use the intake form in resources/requirement_template.md to provide structured input:

Use the AI Development Director skill.
Here is my requirement intake:
[paste filled requirement_template.md]

Mid-project trigger

Use it to validate architecture at any point:

We are 3 weeks into building this app. Use the AI Development Director
to review our current architecture against the original requirements
and identify what risks or gaps exist.

File Reference

File Purpose
SKILL.md Core skill — persona, workflow, rules, output formats
README.md This file — installation, usage, overview
resources/requirement_template.md Structured intake form for new projects
resources/architecture_patterns.md 5 architecture patterns with decision matrix
resources/db_schema_examples.md Schema examples with RLS and index guidance
resources/review_checklist.md Full review checklist with severity ratings
resources/launch_checklist.md Phase-gated launch checklist
examples/sample_input.md Example product idea (AI virtual fitting app)
examples/sample_output.md Full 10-step plan output for the sample input

Who This Is For

  • Solo developers building web apps with AI assistance
  • Small teams using Claude Code or Codex as a development accelerant
  • Founders who need to move from idea to implementation fast without accumulating architectural debt
  • Agencies delivering AI-assisted projects to clients

Who This Is Not For

  • Teams building distributed systems requiring custom infrastructure
  • Projects with dedicated product managers, architects, and DevOps engineers
  • Greenfield enterprise software with compliance-driven architecture requirements

Philosophy

A small, stable system ships. A large, unplanned system does not.

This skill optimizes for: clear requirements → simple architecture → sequential execution → working software.

It does not optimize for: theoretical scalability, architectural elegance, or comprehensive coverage of unlikely edge cases.

License

This product is sold as a commercial digital download.

After purchase, buyers may use, modify, and adapt the included files for personal projects, client work, and internal development workflows.

The purchase price applies to access to this packaged product and its supporting materials.


Similar Agents

More
Discover Agents