UX Wise AI Agent: A Peer-Level Reasoning Skill Pack for Senior Designers
Most AI tools respond to UX questions like a well-read junior designer: they list options, hedge every position, and hand the decision back to you. That is not a thinking partner. That is a search engine with better grammar.UX Wise AI Agent is a structured skill pack that changes how any LLM reasons about design problems, not just what it knows. It installs in under two minutes on Claude Projects, OpenAI Custom GPTs, or any platform that accepts a system prompt. No subscriptions, no accounts, no proprietary platform. You own the files.This is not a prompt list. It is a reasoning framework built on 15 years of AI and product design practice across Microsoft M365 Copilot, Qlik AutoML, and independent AI product work. The skill pack encodes how a senior practitioner actually thinks through hard design decisions, and gives that reasoning posture to the model you already use.What is inside:Three reasoning modes, each behaviorally defined:Strategic Mode is the default. It examines a problem from competing angles, names the trade-offs explicitly, takes a final position, and preempts the strongest counterargument to that position. This is the mode for decisions with downstream consequences you cannot easily reverse.Direct Mode compresses the full reasoning chain into one recommendation and three sentences of rationale. No hedging, no alternatives unless you ask for them. This is the mode for when you need to move.Provocative Mode argues against your stated direction. It finds the assumption doing the most work in your reasoning, puts it under pressure, and closes with a question you have to answer before you proceed. This is not contrarianism. It is the most productive thing you can do before committing to a direction you feel confident about.A knowledge base that operates at practitioner level:All ten Nielsen heuristics documented with applied commentary and diagnostic questions that cut past surface-level compliance to the actual usability problem. Behavioral psychology laws (Fitts, Hick, Miller, Fogg, Kahneman) applied directly to interface decisions. A dedicated AI and UX guide covering trust calibration, automation complacency, explainability requirements, and the ethics of AI-driven interfaces.54 curated starter prompts across five categories, each tagged with the mode that produces the most useful response. Includes a section on why specific inputs produce specific outputs, because the quality of what you get depends entirely on the specificity of what you bring.A vocabulary constraint policy with 35 prohibited words and the reasoning behind each one. "Seamlessly", "actionable insights", "holistic approach", "innovative" as a standalone claim. The words an agent uses signal the quality of the thinking behind them. This policy forces the model out of corporate filler and into specific, defensible language.Full integration guides with working code for the Anthropic API, streaming implementation, multi-turn conversation management, and OpenAI Custom GPT configuration.Three complete worked examples showing what each mode produces on realistic senior-level UX problems. Not hypothetical inputs. Problems that actually show up in enterprise product design.How to activate:Download the files. Open a new Claude Project. Paste the system prompt from prompts/system-prompt.md into Project Instructions. Start with any prompt from starter-prompts.md.That is the entire setup.For API or Custom GPT deployment, follow the guides in the integrations folder.License:MIT. Use it, fork it, extend it, build on it. The GitHub repository is public. This Gumroad page exists for those who want to support the work.Questions:hello@marcosrezende.com
Get it → marcosrezende.gumroad.com