Test an AI product idea before building software: prove the task manually in chat, describe a small agent in a PRD and technical note, then run its instructions against real tools inside Cursor.
Source episode · 40:00
How Dennis Yang Automates Product Management and Prototypes AI with Cursorwith Claire Vo
Dennis turns a morning-briefing PRD into a "super MVP" made of instructions, profiles, and a news-search MCP, then runs the agent directly in Cursor without building a production application.
A no-code agent prototype that completes the target task in Cursor and produces evidence for whether, and how, the team should build it.
AI-first code editor
OpenAI conversational AI
Step by step
Follow the sequence once, then adapt the prompts, checks, and handoffs to your own setup.
4 steps
Run the task manually in a fresh ChatGPT conversation with a representative input. Save the exact prompt, useful output, corrections, edge cases, and what the model could not do without tools.
Produce a [deliverable] for [user] using this input: [input]. Follow these constraints: [constraints]. Cite the source for every factual claim and label missing information instead of filling it in.
In Cursor, turn the learning into a short PRD and a technical note for a "super MVP." Define the job, input, output, tool calls, test data, and excluded production systems.
Write a PRD and technical design for a super MVP of [agent]. It will run inside Cursor using [approved MCP tools] and [synthetic or authorized data]. Define the critical task, input and output schemas, source requirements, failure behavior, evaluation rubric, and test cases. Exclude production authentication, persistence, scheduling, notifications, and deployment.
Write the agent instructions in Markdown. Make the sequence explicit: load configuration, confirm the current date when relevant, call the approved tools, process source material, and return the defined output with citations and uncertainty.
Create agent instructions for this super MVP. Name each allowed tool and its purpose. Require a system-date check, source links and dates, deduplication, a clear no-results response, and a final output that follows [format]. Stop if a required tool or source is unavailable.
Your instructions should be clear and sequential, e.g., '1. Load configuration from profile. 2. Use news_search MCP to find articles. 3. Summarize the content. 4. Generate the final report.'
Run the agent in Cursor across the saved test cases. Record the model, tool calls, output, sources, errors, latency, and cost, then decide whether to revise the task, the instructions, the tool set, or the product idea.
Run today's [agent task] using the super MVP instructions. Use only the configured sources and approved tools. Return the deliverable, source links, unresolved gaps, and a compact execution log.
Turn an idea into a PRD, user stories, and a plan.
After the steps
How to recover when the loop fails and where human judgment helps.
Recover
Keep building
Log your workouts effortlessly without ever opening a spreadsheet. This workflow uses the Claude mobile app to parse a simple voice note and automatically update your workout tracking spreadsheet in the background.
Create a custom email management interface within Claude to read, triage, and draft replies without opening the Gmail tab. This workflow helps you reclaim focus and compound your learnings by training your AI on every interaction.
Build a fully automated system in Claude to generate personalized, interactive HTML proposals and client onboarding materials, saving hours on administrative work and creating a high-touch experience.
Join 100,000+ product managers who use ChatPRD to write better docs, align teams faster, and build products users love.