Skip to content
Aakash GuptaAakash Gupta

Use these skills to supercharge your claude code setup | Oji Udezue | 3x CPO

Most PMs are installing coding skills into Claude Code and stopping there. This episode shows the layer above that, product judgment and business judgment running as skills inside the same repo your engineers use. Oji Udezue scaffolds two products live, watches one clear the viability gate and one get killed by it, then reviews the output the way you would review a PRD in a room full of PMs. Full Writeup: https://www.news.aakashg.com/p/oji-udezue-claude-skills Transcript: https://www.aakashg.com/product-judgment-skills-in-claude-code-with-oji-udezue/ Product Mind Skills Library: https://github.com/ojiudezue/productmind-skills Timestamps: 0:00 - Intro 1:31 - What you will learn today 2:43 - Why PMs became the bottleneck 4:46 - Scaffolding a new product live in Claude Code 8:34 - The viability gate verdict 13:43 - Ads 15:25 - Inside the scaffolding skill step by step 20:15 - Running the gate on a second idea 24:22 - Vet a Feature for existing products 27:25 - Ads 31:18 - Applying PM judgment to the output 37:02 - Standup Zero fails the gate 39:59 - Prototypes and the customer discovery plan 51:59 - The three speed problem 55:31 - Where to grab the skills 🏆 Thanks to our sponsors: 1. Ariso - Ship AI agents and features faster, with fewer regressions - https://ariso.ai/aakash 2. Bolt - Ship AI-powered products 10x faster - https://bolt.new/solutions/product-manager?utm_source=Promoted&utm_medium=email&utm_campaign=aakash-product-growth 3. Product Faculty - Get $150 off their #1 AI PM Certification with code AAKASH150 4. Customer.io - Send smarter messages using your product data - http://customer.io/productgrowth Key Takeaways: 1. Code skills are becoming commodity - GitHub is full of repos that token-max, token-minimize, and cut costs. Every one of them optimizes the same thin slice. Someone already pulled superhuman skills out of their repo because the models got good enough to make them redundant. 2. A tech company runs on three layers - Software and hardware. Product, which is customers and business model. Business, which is where resources get allocated. Almost every skill library on GitHub only touches the first one. 3. PMs are the new bottleneck - Developers with early adopter habits are speeding up fast. Product judgment and orchestration are not speeding up to match. The imbalance shows up as engineers waiting on decisions instead of waiting on code. 4. The viability gate will tell you no - LLMs almost never refuse an idea. This skill scores six dimensions including problem clarity and urgency, target user definition, competitive landscape, differentiation, technical feasibility, and revenue. Three weak scores and it recommends you stop entirely. 5. Watch a real idea get killed - Standup Zero, a Slack-to-standup digest, came back with three moderate scores, a competitive field so crowded it counted against the idea, and thin differentiation. The verdict was proceed with eyes open. Knowing that before you write code is the whole point. 6. LLM market research gets you 60 to 70 percent - Run it, then go find real sources. The failure mode is taking it as gospel. Read the differentiation map and check whether the tools it names are current or decrepit. That is your first credibility scan. 7. Scaffolding is an orchestrator skill - One master SKILL.md calls subskills for market research, the sharp problem test, architecture, testing, and CI. It fetches new skills if your project type needs them. The whole thing runs as an 11-step workflow from business problem to first commit. 8. You get a repo that organizes itself - The skill writes a CLAUDE.md tuned to that specific project, a folder structure that tells both you and Claude what goes where, a test folder, CI so every check-in gets tested, secrets in gitignore, and a bug log that catalogs failures over time. 9. Vet a Feature is the version for existing products - Most PMs are not starting brand new projects. This skill takes one feature idea, tears it apart for anti-patterns, and scores it against the opportunity cost of everything else sitting in your backlog. 10. Shared context beats forked skills - Boris Cherny's CLAUDE.md runs about six lines and points at a shared one. Centralize the skills, harmonize the artifacts with the SDLC your org already expects, and route innovation back into the middle. Ungoverned rollouts kill the compounding. 👨‍💻 Where to find Oji Udezue: LinkedIn: https://www.linkedin.com/in/ojiudezue/ Product Mind: https://www.productmind.co/ Skills Library: https://github.com/ojiudezue/productmind-skills 👨‍💻 Where to find Aakash: Twitter: https://x.com/aakashgupta LinkedIn: https://www.linkedin.com/in/aagupta/ Newsletter: https://www.news.aakashg.com/ #ProductManagement #AI 🧠 About Product Growth: The world's largest podcast focused solely on product + growth, with over 200K+ listeners. 🔔 Subscribe and turn on notifications.

Oji UdezueguestAakash Guptahost
Aug 7, 20261h 4mWatch on YouTube ↗

At a glance

WHAT IT’S REALLY ABOUT

Product judgment skills inside Claude Code to build better, faster

  1. Oji argues that as engineers accelerate with AI, PMs become the new bottleneck unless they also speed up product judgment and orchestration using reusable LLM “skills.”
  2. He demos a “New Project Scaffolding” orchestrator skill that runs an 11-step workflow: viability gating, market research, product brief creation, repo setup, testing, CI/CD, security hygiene, and a tailored CLAUDE.md.
  3. The viability gate is positioned as a rare mechanism that can credibly say “no” to weak ideas using explicit PM frameworks, not just agreeable chat behavior, and it is tested on two ideas with different outcomes.
  4. Oji emphasizes that LLM outputs are drafts, not truth—PMs must review artifacts (market research, PRD/brief) for credibility, missing personas, and strategic fit before proceeding.
  5. He extends the workflow beyond ideation into prototypes and customer discovery plans, and closes with guidance for rolling these skills out in teams via shared context and centrally governed CLAUDE.md/skill libraries.

IDEAS WORTH REMEMBERING

5 ideas

AI makes PM judgment the limiting factor unless it’s systematized.

Oji observes engineers speed up rapidly with AI, while PM workflows (framing, prioritization, validation) often stay manual; codifying judgment into repeatable “skills” helps PMs operate at engineering speed.

Treat product building as a three-layer stack: business, product, and code.

He argues most AI tooling over-focuses on coding efficiency, but successful products require viability (business), discovery and UX (product), and quality delivery (code) working together in one flow.

A viability gate that can say “no” is a practical antidote to LLM agreeableness.

The scaffolding skill uses explicit eval dimensions (e.g., problem clarity, target user, competition, differentiation, feasibility, revenue) and recommends stopping when multiple dimensions are weak—something generic chat rarely does.

Scaffolding is not just boilerplate—it’s a handoff and quality system for non-engineers.

Beyond docs, the workflow creates a structured repo, a tailored CLAUDE.md, tests, CI templates, and security safeguards (e.g., secret hygiene), reducing the common “vibe-coded but unshippable” gap.

Market research from LLMs is useful only with credibility checks and follow-up validation.

Oji recommends using LLM research for 60–70% of initial coverage, then reviewing for staleness, missing competitors, and internal consistency before trusting conclusions or investing further.

WORDS WORTH SAVING

5 quotes

The era of mono-skilled professionals is dead, so I'm going to demonstrate the future of product builders

— Oji Udezue

I've been a product manager for 25 years, and I've never really felt that coding was worth my time until now.

— Oji Udezue

LLMs very rarely tell you no, and this skill will tell you absolutely no, don't do this.

— Oji Udezue

The danger in, in this kind of workflow is that you take it as gospel, right? You really do need to look through all the output, make sure that it makes sense for you.

— Oji Udezue

There's something we call the three-speed problem, which is for the past forty, fifty years, whatever you wanna call it, the longest pole has always been, uh, development.

— Oji Udezue

PMs as bottleneck in AI-accelerated teamsSkills/orchestrators in Claude Code (multi-skill workflows)Viability gate vs. sharp problem test frameworksMarket research + product brief generation in-repoRepo scaffolding: CLAUDE.md, folder structure, CI/CD, testing, securityPrototyping interaction models with design skillsCustomer discovery week plan and recruitment gatingThree-speed problem: why/build/distribute moving at different ratesFeature evaluation: “Vet a Feature” skillEnterprise rollout pitfalls: forking, governance, shared context

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

iOSAndroidClaudeChrome