AI Agent Spec Claude Skill — System Prompt & Architecture Template

Specify production AI agents end-to-end: persona, tools, memory, guardrails, evals, and rollout. The Claude Skill agency teams use to ship agents that actually work in production — not just demos.

What this Claude Skill does

• Define agent persona, tools, memory, and behavioral constraints in one structured spec
• Lock in evals and red-team scenarios before launch so quality stays high
• Map handoffs to humans and to other agents so the system doesn't get stuck
• Document deployment, monitoring, and rollback so the agent stays reliable post-launch

Built for

• AI engineers building production agents on Claude, OpenAI, or open-source LLMs
• Founders writing the spec before handing off to engineering
• Agencies productizing agent builds for clients

What's inside

• SKILL.md — the full skill file with frontmatter, methodology, phases, examples, anti-patterns, and recovery
• Step-by-step phases Claude follows to produce a senior-level output every time
• Worked examples with realistic detail you can adapt to your situation
• Anti-patterns + recovery playbook so the work survives edge cases
• Delivery file tree so outputs land organized for you or your client

How a Claude Skill works

This is a Claude Skill — a single SKILL.md file Claude reads as instructions for a specific kind of work. Drop it in your Claude Project, your /mnt/skills/user/ folder, or wherever Claude can read it. Claude triggers the skill automatically when your prompt matches its description, then follows the methodology end to end.

Author

Built by Ukiyo Productions — a strategy-led branding and growth agency. Part of a 100+ skill marketplace built for founders, operators, consultants, and agencies.