AI Agent Process — AI skill for Claude Code

AI Agent Process

A structured workflow framework for AI-powered development with Claude Code. Provides role separation, iteration management, and decision frameworks that turn ad-hoc AI coding into a repeatable development process.

**Philosophy:** Ship pragmatically, iterate deliberately, pivot when you learn.

---

This Framework Is Not About AI Doing Everything

Let's be clear: **this is not an autopilot.**

The goal is not to type a prompt and walk away while AI builds your product. That path leads to mediocre software, missed edge cases, and systems nobody understands. AI can write code fast, but speed without direction produces noise.

This framework exists to **elevate your work, not eliminate it.** You're still the architect. You still make the hard calls. You still own the outcome. What changes is *where* you spend your energy:

• You define what matters — requirements, acceptance criteria, quality standards
• You make judgment calls — when to pivot, what tradeoffs to accept, when "good enough" is good enough
• You review with intention — not rubber-stamping AI output, but genuinely evaluating whether it solves the problem
• You build understanding — the knowledge base, the patterns, the "why" behind decisions

AI agents handle the mechanical work: writing boilerplate, running validations, organizing artifacts, executing repetitive tasks. They're tireless and consistent. But they don't know what your users need. They don't feel the weight of technical debt. They can't tell when a "working" solution misses the point entirely.

**The human-in-the-middle isn't a bottleneck — it's the whole point.**

Great software comes from humans who care deeply about the problem, supported by tools that handle the grunt work. This framework structures that collaboration: AI does the coding and scope organization, you do the thinking and deciding. The result isn't "AI-generated code" — it's *your* work, built faster and more consistently.

If you want t