prompt-master

prompt-master generates optimized, production-ready prompts for a wide range of AI tools — LLMs like Claude, ChatGPT, and Gemini, reasoning models, coding agents such as Cursor, and image and video generators like Midjourney. It solves the problem of turning a rough idea into a single copy-paste-ready prompt tuned to the quirks of the specific target tool, activating only when the user explicitly asks to write, fix, improve, or adapt a prompt and staying out of general conversation and coding tasks.

The skill operates as a disciplined prompt engineer. Before writing anything it silently extracts nine intent dimensions (task, target tool, output format, constraints, input, context, audience, success criteria, examples), asking at most three clarifying questions and never producing a prompt without first confirming the target tool. It routes to tool-specific guidance — for example front-loading and literal specificity for Claude 4.x, minimal short instructions with no chain-of-thought for reasoning-native models like o3 and DeepSeek-R1, and compact structured outputs for GPT-5.x. Hard rules steer toward simpler techniques (role assignment, few-shot, grounding anchors) and flag higher-fabrication-risk meta-reasoning frameworks such as Tree of Thought or Mixture of Experts as opt-in only. Output is locked to a single copyable prompt block plus a one-line note on what was optimized, drawing on bundled pattern and template reference files.

The skill targets anyone who works with AI tools and wants better results — developers, content creators, and prompt engineers. It is pure text-generation guidance with no code execution, network calls, or credential handling.