← ClaudeAtlas

system-prompt-builderlisted

Interview-driven builder for enterprise-grade AI system prompts and agent architectures. Use whenever the user wants to design, build, scaffold, or rebuild a system prompt, custom assistant, AI agent, project instructions, custom GPT, multi-agent system, or any reusable LLM instruction set. Trigger on phrases like "build me a prompt system", "design an agent", "write system instructions", "turn this into an agent", "create project instructions", "I want an AI that does X reliably", or any request to architect how an AI should behave across many future interactions — even when the user never says the words "system prompt." Also trigger when the user pastes an existing system prompt and wants it rebuilt, hardened, or extended.
scoobydrew83/skills · ★ 0 · AI & Automation · score 78
Install: claude install-skill scoobydrew83/skills
# System Prompt Builder Build production-grade system prompts and agent architectures by interviewing the user, then generating a complete, ready-to-paste instruction set. This skill does not hand back generic templates — it asks the right questions, enforces a set of hard-won design principles, and produces a system prompt the user can deploy as-is. The output is always a **complete, self-contained system prompt as a `.md` file**, optionally followed by a **review pass** that scores the result and names the single highest-impact improvement. --- ## The four archetypes Every system you'll build falls into one of four shapes. Each has a different skeleton, a different interview, and different failure modes. Your first job is to classify, because everything downstream depends on it. | Archetype | Build when the user wants… | Reference file | |-----------|----------------------------|----------------| | **A — Mode-based assistant** | A reusable assistant/tool that handles several distinct request types (commands/modes), each with its own output shape. *(e.g. a prompt-engineering helper with BUILD / REVIEW / GENERATE modes)* | `references/archetype-mode-based-assistant.md` | | **B — Phased human-in-the-loop agent** | An agent that researches/analyzes/plans across sequential phases, pausing for human confirmation, while the human keeps decision authority | `references/archetype-phased-agent.md` | | **C — Single-purpose task prompt** | One focused, repeatable instruction for