15 Prompts That Work Better When Your AI Already Knows Your Product
I have 71 prompts in the Prompt-Led Vault. These are the 15 a Product Manager should run first, and why every single one changes the moment Claude knows your context.
I am going to tell you something that annoyed me for a while before I fixed it.
Last week I published the Product Brain MCP tutorial. Many of you built the vault. You had the folder, the script, the hammer icon. And then a few of you messaged me with some version of the same question:
“Okay. Now what do I actually ask it?”
That’s fair! Building the bridge is one thing. Knowing which trips to take across it is another.
Here is the honest answer. I have 71 prompts inside the Prompt-Led Vault. Some are for beginners. Some are for PMs who are already shipping apps. Some are free. Some are paid. And most of them are fine when Claude is working from memory alone.
But 15 of them become genuinely different tools when Claude already knows your product. Not slightly better. Different. The output shifts from generic best-practice advice to something that reads like it was written by someone who was in your last sprint review.
I’ll guide you through this gap now.
The roadmap for today’s deep dive:
The contrast that changes everything: Without vault vs. with vault, shown with a real example.
The free five: The prompts to run this week, even if you are just getting started.
The paid ten: The prompts that go deeper once Claude knows your product cold.
How to use these together: The sequence that turns 15 prompts into a full PM workflow.
Let’s get into it. 👇
The Contrast That Changes Everything
Before we get into the list, I want to show you exactly what I mean. Because this is the thing most prompt libraries do not cover. A prompt library is a list. A vault-connected prompt library is a system.
For example, I had a decision sitting in my roadmap from a few weeks ago:
Guests invited to a DraftKit.app collaboration session should be able to see the creator’s calendar and pitch without creating an account. The sign-up wall only appears when we move into the workspace together.
The reasoning was documented: forcing a guest to register before they see any value taxes the inviting creator’s reputation, not just the guest’s patience.
That decision was in a spec file called draftkit-guest-flow.md sitting in my vault.
Here is what happens when I ask the same prompt two ways.
The prompt: “Write a user story for the guest onboarding flow.”
❌ Without the vault:
Claude has no idea what DraftKit is. It writes from best practices.
“As a guest user, I want to be able to join a collaboration session, so that I can work on content with the creator.”
Technically a user story. Completely disconnected from the actual design decision already made. No mention of the account wall. No mention of the calendar view. No mention of when the sign-up trigger should fire. Even a developer reading this would build the wrong thing.
✅ With the vault:
Claude reads draftkit-guest-flow.md before responding. It knows the decision. It writes to the constraint.
“As a guest invited to a DraftKit session, I want to view the creator’s pitch and availability calendar before I am asked to create an account, so that I understand the collaboration context and the creator’s reputation is protected by not front-loading friction on people they personally invited.”
Acceptance criteria:
Guest can view creator calendar and active pitch without authentication
Sign-up prompt appears only when guest clicks “Enter Workspace”
Guest session expires after 48 hours if no workspace action is taken
Same prompt. Three seconds of difference in setup. Completely different output.
The second story does not need a meeting to explain. A developer can build directly from it. A stakeholder can validate it. And it matches a decision you already made weeks ago without you having to remember it or type it again.
That is what the vault does.
The Free Five: Start Here This Week
These five prompts are available to all subscribers. They cover the core of a PM’s weekly workflow. Run them in order and you will have a cleaner week than you started with.
1. The Product Brain Logic Vault
Difficulty: Intermediate | Category: Product Insight | 🆓 Free
Stop being the team librarian. In 2026, if your product’s history only exists in your head and Slack threads, you are the single point of failure.
The problem it solves: You are wasting the first 15 minutes of every AI session re-explaining project history. This prompt converts your AI from a stateless chatbot into a Product Brain that retains context across sessions.
❌ Without the vault: Claude asks you to explain your product from scratch every time. You paste context. It responds. You start the next session and do it again.
✅ With the vault: Claude reads your specs before it responds. You ask a question. It already knows the backstory.
The prompt:
Role: Senior Product Manager
Context: I have a Product Brain MCP connected to a local
`product-vault` folder containing my product specs.
The MCP has a tool called `read_spec(filename)`.
Task:
1. Read the following specs from my vault:
`[spec_name_1.md]` and `[spec_name_2.md]`.
2. Based on these documents, identify the top 3 technical
risks for the upcoming `[feature_name]` launch.
3. Draft a pre-mortem document outlining a mitigation
strategy for each risk.
Input: The connected `product-vault`.
Strategic advice: Stop acting like a human database. In 2026, your value is not remembering decisions. It is building systems that do. If you are manually bridging information gaps, you are the bottleneck.
This is the prerequisite for everything else on this list. If you have not built it yet, start here.
2. The Intent-Driven User Story Framework
Difficulty: Beginner | Category: Method | 🆓 Free
User stories are the syntax of your product’s soul. In 2026, if you cannot describe the value in one sentence, you have not architected the logic.
The problem it solves: Your development team, human or AI, is building features that do not solve real user friction. Traditional user stories focus on “The What.” This framework focuses on “The Why” and “The Result.” It creates Intent Units that are clear enough for an AI builder to execute and a stakeholder to validate.
❌ Without the vault: Claude gives you a generic three-part story template. Fine as a starting point, nothing more.
✅ With the vault: Claude reads your existing acceptance criteria format, checks the user type taxonomy you have documented, and writes a story that matches your team’s actual language and your product’s specific constraints. It will also flag if the new story contradicts something you already decided.
The prompt:
Role: Technical Product Architect / Story Writer.
Context: I am drafting a set of user stories
for [Insert Feature Name].
Task:
1. Generate 3 user stories using the
"As a... I want to... So that..." format.
2. For each story, provide 3 Acceptance Criteria
that include specific technical logic gates
(e.g., "If user is unpaid, show upgrade modal").
3. Identify the Success Metric for each story
that proves the user friction has been removed.
Input: [Insert high-level feature idea and target persona].
Strategic advice: Your story is the code before the code exists. A user story is not a suggestion. It is a specification of intent. If you leave it open to interpretation, you are inviting the AI to guess. Do not describe the button. Describe the decision the user makes when they see the button.
3. The High-Signal Weekly Update
Difficulty: Beginner | Category: Process | 🆓 Free
Stakeholders do not want a list of tasks. They want a pulse check on your technical intent. In 2026, if your update requires a meeting to explain, it is just noise.
The problem it solves: You are spending hours every Friday manually drafting status reports that stakeholders skim or ignore. This prompt synthesizes technical progress into a high-signal executive summary that focuses on logic milestones and risk mitigation, not tasks completed.
❌ Without the vault: Claude writes a generic weekly update template. You fill in the blanks. It sounds like every other PM update in every other company.
✅ With the vault: Claude reads your last sprint notes, cross-references your roadmap, and writes an update that specifically names the decisions made this week, the risks surfaced, and the blockers that need executive attention before Monday.
The prompt:
Role: Strategic Communications Lead / AI PM.
Context: I am drafting a high-signal weekly update
for stakeholders based on the following raw data:
[Insert Jira Export / Slack Summary].
Task:
1. Identify the 3 most significant Technical Wins
that directly impact business ROI.
2. Summarize 1 Logic Risk we encountered
and how we mitigated it.
3. Draft a 3-paragraph update using a TL;DR format
that focuses on Intent, Velocity, and Next Steps.
Input: [Insert raw weekly logs and current product goals].
Strategic advice: Your update is a product, not a chore. The PM who wins provides the highest signal-to-noise ratio. If your stakeholders have to ask “What does this mean for the business,” you have failed the update. Use AI to do the reporting so you can do the reframing.
4. The High-Velocity Backlog Refinement
Difficulty: Intermediate | Category: Process | 🆓 Free
In 2026, your backlog is either the fuel for your next sprint or the graveyard of your focus. Stop grooming and start architecting intent.
The problem it solves: Your backlog has become a bloated list of “someday” features and your refinement sessions feel like a repetitive chore. This prompt moves you from Ticket Management to Logic Orchestration by ensuring every item is a functional vertical slice before it enters a sprint.
❌ Without the vault: Claude helps you rewrite ticket titles and add acceptance criteria. Useful, disconnected from your actual product decisions.
✅ With the vault: Claude reads your existing roadmap priorities, flags any ticket that contradicts a decision already made, and reorders the backlog based on the technical dependencies it can see in your specs. It also marks tickets as “Not Logic-Ready” rather than letting them through.
The prompt:
Role: Backlog Architect / Technical Lead.
Context: I am refining a product backlog
for [Insert Product / Project].
Task:
1. Analyze the following backlog items: [Insert List].
2. Identify 3 Duplicate or Low-Utility items that
should be deleted to preserve team focus.
3. For the top 3 items, generate a technical Logic Spec
that a developer can use to build a functional
prototype in under 2 hours.
4. Rank the remaining items by Architectural Alignment
with our 2026 goals.
Input: [Insert raw backlog items and current product focus].
Strategic advice: A clean backlog is a psychological safety net. When a developer sees 500 open tickets, they see a mountain they can never climb. When they see 10 Logic-Ready slices, they see a roadmap they can ship. The PM who wins in 2026 says no to the backlog 80% of the time.
5. The Vertical Slicing Framework
Difficulty: Intermediate | Category: Method | 🆓 Free
Stop building layers and start shipping value. In 2026, a “Back-end Only” sprint is a signal of technical inertia.
The problem it solves: Your team is taking weeks to “lay the foundation” without shipping a single functional screen. Horizontal slicing creates a high-risk integration period at the end. This framework breaks features into Vertical Slices that deliver end-to-end utility from day one.
❌ Without the vault: Claude explains what vertical slicing is and gives you a theoretical example with a fictional product.
✅ With the vault: Claude reads your current feature list, identifies which items are actually horizontal layers pretending to be features, and proposes a re-sliced version of your next three sprints that ships something clickable every week.
The prompt:
Role: Technical Product Manager / Agile Orchestrator.
Context: I am breaking down [Insert Feature Name]
into vertical slices.
Task:
1. Identify the 3 smallest Vertical Slices that provide
end-to-end value for the user.
2. Define the database schema, API endpoint requirements,
and UI components for Slice 1.
3. Propose a Refinement Sequence that allows us to ship
Slice 1 to production while the other slices
are still in development.
Input: [Insert Feature PRD or high-level requirements].
Strategic advice: The slice is the unit of progress. In 2026, “Percentage Complete” is a vanity metric. The only real metric is “Percentage Shipped.” If you cannot click through the logic from UI to database, you have not built anything yet.
The Paid Ten: Go Deeper Once Claude Knows Your Product
These ten prompts are available to paid subscribers. They assume you have already built the vault, already run the free five at least once, and now want to use Claude as a senior thought partner, not just a writing assistant.
Each one gets sharper the more context your vault contains. If you have your roadmap, your stakeholder notes, your release history, and your acceptance criteria all in the vault folder, these prompts produce output that is almost unfair.
6. The Product Refinement Truth-Check
Difficulty: Intermediate | Category: Audit | 🔒 Paid
Refinement is not a reading session. It is a technical stress-test. In 2026, if you are not deleting tickets during refinement, you are not refining anything.
The problem it solves: Your refinement sessions have become a boring walkthrough of Jira tickets everyone has already seen. This audit turns refinement into a high-stakes logic review where the goal is to find the Single Point of Failure in every user story before development begins.
❌ Without the vault: Claude reviews your ticket text and suggests improvements to the acceptance criteria. It still cannot tell you if this ticket contradicts a decision made three sprints ago.
✅ With the vault: Claude reads the ticket, cross-references it against your decision log, and flags any logical conflict, missing edge case, and unvalidated assumption. It marks tickets as Logic-Ready or Not Logic-Ready before your developers ever see them.
The prompt:
Role: Technical Product Lead / Agile Orchestrator.
Context: I am leading a refinement session and need to
Truth-Check our current sprint candidates.
Task:
1. Review the following 3 user stories: [Insert Stories].
2. Identify the Vague Logic in each story that could
lead to technical debt or AI-builder errors.
3. Re-draft the Acceptance Criteria to include 3 specific
Technical Hard-Gates
(e.g., "User cannot proceed if API response > 200ms").
4. Rank these stories by Implementation Certainty.
Input: [Insert raw user stories and current sprint goals].
Strategic advice: Your refinement is a filter, not a funnel. If you let fuzzy logic through refinement, you are pre-ordering a bug for next week. Be the PM who is brave enough to say “This is not ready for the team.”
7. The Stakeholder Resonance Trick
Difficulty: Intermediate | Category: Method | 🔒 Paid
Stakeholders do not buy roadmaps. They buy reality. Stop asking for permission and start showing the prototype.
The problem it solves: You are stuck in a cycle of endless meetings and conflicting feedback. Logic debates are subjective. Functional prototypes are objective. This prompt uses Proof of Intent to collapse stakeholder misalignment in a single demo session.
❌ Without the vault: Claude helps you prepare for a stakeholder meeting with generic talking points and slide structure suggestions.
✅ With the vault: Claude reads your stakeholder communication history, identifies the specific objections that have come up before, and builds a demo flow that addresses each one with a functional prototype moment. It knows what your VP cares about because it has read the notes from the last four conversations with them.
The prompt:
Role: AI Product Architect / Stakeholder Strategist.
Context: I am building a Proof of Intent prototype
to win over a skeptical stakeholder
regarding [Insert Feature].
Task:
1. Analyze the stakeholder's primary objection:
[Insert Objection].
2. Design a single-page prototype that solves
this specific objection.
3. Draft 3 technical Talking Points that anchor
the demo in functional reality rather than
aspirational theory.
Input: [Insert Objection and Feature Specs].
Strategic advice: The demo is the decision. In 2026, the PM who cannot build is just a narrator of other people’s work. A working prototype is the only stakeholder management tool that does not require a slide deck.
8. The High-Velocity Discovery Engine
Difficulty: Intermediate | Category: Method | 🔒 Paid
In 2026, discovery is not a phase. It is a continuous technical requirement. If you are not talking to users, you are just hallucinating your roadmap.
The problem it solves: You are about to commit engineering resources to a new initiative but lack functional evidence that it solves real user friction. This engine bridges a Good Idea and a Proven Utility through rapid prototyping, user synthesis, and hypothesis testing.
❌ Without the vault: Claude gives you a discovery interview template and a hypothesis canvas. Neither knows anything about what you have already tried.
✅ With the vault: Claude reads your previous discovery notes, flags patterns you missed, and surfaces contradictions between what users said six months ago and what stakeholders are now asking for. It tells you which hypotheses you already tested so you do not run the same experiment twice.
The prompts (run in sequence):
[PROMPT 1: THE HYPOTHESIS GENERATOR]
Review the project brief for [Insert Project].
Identify the 3 biggest Risks (User, Business, Technical).
Draft a hypothesis for each in the format:
"We believe [User] will [Action] because [Reason]."
[PROMPT 2: THE DISCOVERY SYNTHESIZER]
Analyze the following user interview notes: [Insert Notes].
Identify 3 recurring themes of Friction and
2 Unexpected Use Cases.
Rank these by their impact on our core metric:
[Insert Metric, e.g., Onboarding Success].
[PROMPT 3: THE EXPERIMENT DESIGNER]
I need to validate [Hypothesis].
Design a low-cost, 48-hour experiment that provides
Functional Evidence of user desirability.
Include the Success Threshold required to move
this to a full build.
Strategic advice: Discovery is the ultimate technical filter. Do not ask users what they want. Watch what they struggle with. If the evidence does not exist, the feature does not get built.
9. The Technical Clarity Playbook
Difficulty: Intermediate | Category: Playbook | 🔒 Paid
The era of the non-technical PM is over. In 2026, technical clarity is your only leverage.
The problem it solves: Your engineering team is hitting hallucination debt because your requirements are too vague. High-agency PMs do not just describe features. They architect the logic gates that make the product possible.
❌ Without the vault: Claude rewrites your requirements to be more specific. Useful for a single ticket, does not change your overall clarity as a communicator.
✅ With the vault: Claude audits your last five requirements documents for recurring vagueness patterns. It tells you specifically which types of decisions you tend to leave unresolved and gives you a personal clarity protocol based on your own blind spots, not a generic template.
The prompt:
Role: AI Product Architect.
Context: I am moving from a descriptive PRD
to a technical logic spec.
Task:
1. Analyze [Insert Feature Idea] and identify the
3 most critical Technical Logic Gates (if/then paths).
2. Write a system prompt for a Functional Ghost
that executes this logic.
3. Identify the potential Hallucination Debt risks
where the AI might deviate from my product intent.
Input: [Insert Feature Draft].
Strategic advice: Clarity is the new code. The person who defines the architecture wins. Stop writing user stories and start writing logic gate specs.
10. The Dual-Value Prioritization Matrix
Difficulty: Intermediate | Category: Framework | 🔒 Paid
If a feature solves a user problem but does not move a business metric, it is a hobby. If it moves a metric but creates user friction, it is technical debt. Build where they overlap.
The problem it solves: Your roadmap feels like a wish list. This framework forces a clinical assessment of every roadmap item against two axes: User Desirability and Business Viability. You only build high-leverage assets that serve both.
❌ Without the vault: Claude applies the matrix to a feature you describe in the chat. Accurate but hypothetical.
✅ With the vault: Claude reads your full roadmap, applies the matrix to every item simultaneously, and produces a ranked list with reasoning for each placement. Items that score low on both axes get flagged for deletion, not deprioritization.
The prompts (run in sequence):
[PROMPT 1: THE USER-VALUE AUDIT]
Analyze [Feature Name] for our [Target Persona].
Identify the specific Friction Point it removes.
On a scale of 1-10, how much does this improve
their Daily Active Utility?
Provide evidence-based reasoning.
[PROMPT 2: THE BUSINESS-VALUE AUDIT]
Evaluate [Feature Name] against our current
business goals: [Insert Goals, e.g., Reduce Churn 5%].
What is the projected ROI over 6 months?
How does this align with our long-term
market positioning?
[PROMPT 3: THE MATRIX SYNTHESIZER]
I have a list of features with User and Business scores:
[Insert Data].
Propose a prioritized roadmap for the next 4 weeks
that maximizes ROI while maintaining a high-fidelity
user experience.
Strategic advice: Your roadmap is a statement of intent. The PM who wins says no to a good idea to make room for a great one. If you cannot prove the ROI and the User Win, do not build it.
11. The Developer Risk Disclosure Audit
Difficulty: Intermediate | Category: Audit | 🔒 Paid
Silence is a technical risk. In 2026, if your developers are not telling you why the AI will fail, it is because they do not feel safe enough to be right.
The problem it solves: Your development velocity is high but your bug count is quietly exploding. In AI building, silence often hides brittle prompts, unvetted training data, or security shortcuts. This audit rebuilds the Technical Truth Loop between Product and Engineering.
❌ Without the vault: Claude gives you a list of questions to ask your developers in your next 1:1.
✅ With the vault: Claude reads your recent release notes and incident logs, identifies patterns in where bugs are clustering, and prepares a targeted risk disclosure conversation that addresses the specific parts of your codebase showing stress signals.
The prompt:
Role: Technical Culture Auditor / AI Lead.
Context: I am performing a Risk Culture Check to ensure
developers feel empowered to flag AI vulnerabilities.
Task:
1. Review the last 3 AI feature launches and identify
every Technical Concern that was raised but not fixed.
2. Identify the Hidden Debt developers are afraid
to mention because of current roadmap pressure.
3. Propose a Safety-First communication channel where
risks can be flagged without impacting personal
velocity metrics.
Input: [Insert current roadmap and recent bug reports].
Strategic advice: Velocity without transparency is a suicide mission. Create a culture where a developer can say “The AI is lying to us” without being seen as a blocker. Stop celebrating speed. Start celebrating honesty.
12. The AI Product Rescue Manual
Difficulty: Intermediate | Category: Playbook | 🔒 Paid
In 2026, burnout is a signal that your process is manual, not that your workload is high. Use AI to reclaim your cognitive bandwidth.
The problem it solves: You are drowning in low-value execution: competitor tracking, user feedback tagging, drafting boilerplate docs. This manual provides eight specific Technical Rescues that automate the administrative overhead of product management so you can focus on architectural intent.
❌ Without the vault: Claude gives you a list of tasks you could theoretically delegate to AI.
✅ With the vault: Claude reads your typical week from your notes and process docs, identifies the three tasks consuming the most time with the lowest decision value, and builds the exact prompts to automate each one with your specific context already baked in.
The prompts:
[PROMPT: THE COMPETITOR SYNTHESIZER]
Analyze the latest feature updates for [Competitor A]
and [Competitor B].
Identify the 3 logic gates they have implemented
that we lack, and rank them by competitive threat
to our [Metric, e.g., Retention].
[PROMPT: THE PRD BOILERPLATE]
Draft a technical PRD for [Feature Name] using a
logic-first structure.
Define the Core Utility, the 5 primary User Flows,
and the Success Metrics for the first 48 hours of launch.
[PROMPT: THE FEEDBACK AGGREGATOR]
Review the following 50 user feedback entries:
[Insert Text].
Categorize them into technical bugs, UX friction,
and feature requests.
Identify the Single Point of Failure
mentioned most frequently.
Strategic advice: Your value is in the decisions, not the drafts. Every minute you spend manually summarizing a meeting is a minute you are not spending on product strategy.
13. The 2026 Capability Roadmap Blueprint
Difficulty: Intermediate | Category: Blueprint | 🔒 Paid
A 12-month feature list is a work of fiction. In 2026, you do not commit to features. You commit to technical capabilities.
The problem it solves: Your roadmap feels like a heavy list of promises you will likely break. AI velocity makes static planning obsolete. This blueprint shifts you to a fluid architecture that responds to market resonance instead of quarterly planning cycles.
❌ Without the vault: Claude builds you a capability roadmap template. Looks good in a slide. Disconnected from what your product actually does today.
✅ With the vault: Claude reads your existing roadmap and specs, identifies the technical capabilities you already have but are not fully exploiting, and proposes a Q2 through Q4 arc that builds on proven foundations rather than starting from scratch every quarter.
The prompt:
Role: AI Product Architect.
Context: I am refactoring a static 2026 roadmap
into a capability-based blueprint.
Task:
1. Analyze the current roadmap: [Insert Roadmap].
2. Convert 3 Static Features into Dynamic Capabilities.
3. Define the technical Logic Gates required to unlock
the next phase of growth.
Input: [Insert current roadmap or feature list].
Strategic advice: Fixed dates are the enemy of technical velocity. The best roadmap is a series of logic gates. If you cannot explain the capability behind the feature, you are guessing. Architect the power and the features will follow.
14. The Brutal Auditor Logic
Difficulty: Intermediate | Category: Audit | 🔒 Paid
Your bias is a bug. If you are not using AI to find the 12 ways your app will fail, your users will find them for you.
The problem it solves: You are too close to your product to see the friction. Most builders assume users will follow the Happy Path. This logic forces the AI to hunt for the Angry Path and the Edge Case failures before you launch.
❌ Without the vault: Claude stress-tests a feature description you write from memory. It finds generic failure modes that apply to any product.
✅ With the vault: Claude reads your actual spec, your acceptance criteria, your past incident reports, and your user feedback logs. Then it hunts for failure modes specific to your product’s architecture, your users’ actual behavior patterns, and the technical decisions you have already locked in. It finds bugs that are already growing inside your current build.
The prompt:
Role: Hostile UX Auditor / Senior Technical Architect.
Context: I am submitting my app logic for a Brutal Audit.
Task:
1. Review [Insert Product URL / Logic / Screenshots].
2. Find 12 major flaws including UX friction,
logic loops, and potential database dead-ends.
3. For every flaw, suggest a Technical Fix that can be
implemented in under 2 hours.
Input: [Insert App Details / Logic / Screenshots].
Strategic advice: High-agency building requires high-fidelity feedback. If you only ask the AI to “make it better,” it will lie to you. You must command it to tear it apart. Do not defend your code. Fix the intent.
15. The 2026 Technical PM Blueprint
Difficulty: Intermediate | Category: Blueprint | 🔒 Paid
The role of the non-technical PM is extinct. In 2026, you do not manage developers. You architect intent.
The problem it solves: You feel stuck in the Coordination Trap: spending all your time in meetings instead of building. This blueprint defines the shift from project manager to AI product architect, focused on high-agency technical leadership.
❌ Without the vault: Claude gives you a framework for becoming a more technical PM. Good reading. Nothing you cannot find in a blog post.
✅ With the vault: Claude reads your actual work history from your builder log and decision docs, identifies the specific gaps between where you are operating now and where a Technical PM operates, and gives you a personalized 90-day plan using your current projects as the training ground. It is a diagnosis of your specific situation, not a career template.
The prompt:
Role: AI Product Architect / Strategic Consultant.
Task: Audit my current Product Management workflow:
[Insert Workflow / Tasks].
1. Identify 3 areas where I am acting as a Reporter
instead of an Architect.
2. Suggest a Technical Logic Spec template
for my next feature: [Insert Feature].
3. Define the Technical Velocity metrics I should track
to prove I am building, not just managing.
Input: [Insert current job description
or weekly task list].
Strategic advice: Architecture is the new leadership. In an AI-saturated market, the value of management is approaching zero. Your edge is your ability to precisely define the technical constraints and logic that the AI executes. Do not be the person who checks the status. Be the person who defines the architecture.
How to Connect Your Claude to the Vault Directly
Every prompt in this post was written to work inside Claude. But when Claude can search the full Vault library without you ever leaving the chat, the whole system changes.
Instead of opening a tab, searching, copying, and pasting, you just ask:
“Search the vault for a prompt that helps me prioritize my roadmap.”
Claude finds it, loads the full logic, and you run it immediately. No tab switching. No copy-pasting. Just the prompt, ready when you need it.
This is available to paid subscribers. Here is the setup, three steps, no coding required.
Step 1. Download Claude Desktop from claude.ai/download if you do not have it yet. Sign in with your Anthropic account.
Step 2. Log into your account at promptledproduct.com/account. Click “MCP Access Key” in the Developer section. Hit “Generate MCP Key” and copy what appears.
Step 3. In Claude Desktop, open Settings → Developer → Edit Config. Paste the config with your key replacing YOUR_KEY_HERE. Save the file.
{
"vault-search": {
"url": "https://jnhsksmhrelwwrtdrapx.supabase.co/functions/v1/vault-search",
"method": "POST",
"headers": {
"x-vault-api-key": "YOUR_KEY_HERE",
"Content-Type": "application/json"
}
}
}⚠️ Two things that will catch you out: MCP tools only work in Claude Desktop, not in the browser version at claude.ai. They also only appear in regular chats, not inside Projects. Start a new chat from the Claude Desktop home screen, not from inside a project.
Step 4. Fully quit Claude Desktop with Cmd+Q and reopen it. Then go to Customize → Connectors, find vault-search, and set it to Always Allow. Claude will not use the tool until you do this.
I always try to make things easier for my users. So instead of copying the key and creating the file yourself, you can download the file you need to replace it.
You are connected to all 71 prompts.
Your Vault MCP key is waiting. Log into your account at promptledproduct.com/vault, click Developer, and it is there. If you are not a paid subscriber yet, this is the most concrete thing I can show you for why it is worth it. You connect it in three minutes and you never copy-paste a prompt again.
How to Use These 15 as a System, Not a List
Here is the sequence I would run if I were starting fresh today.
1️⃣ Day one. Build the vault if you have not already. Drop in your roadmap, your last three sprint notes, and your most recent stakeholder email thread. That is enough context to start.
2️⃣ Day two. Run prompts 2, 3, and 4. Write one Intent-Driven story, draft your weekly update, and clean your backlog in one session. That is a full productive morning.
3️⃣ End of week one. Run prompt 5 on your next sprint. Identify which items are horizontal layers pretending to be features. Delete two of them.
4️⃣ Week two onward. Add the paid prompts one at a time, starting with 14 (Brutal Auditor) and 10 (Dual-Value Matrix). Those two together will change how you look at your roadmap permanently.
The vault gets smarter as you add more to it. Every spec you write, every decision you document, every stakeholder note you save makes the next prompt sharper. The system compounds.
Where Product Management Is Actually Going
Here is the uncomfortable truth behind all 15 of these prompts.
The PM who runs them without a vault gets generic output. Useful, forgettable, replaceable by anyone with access to the same model.
The PM who runs them with a vault gets output that knows their product, their team’s language, their previous decisions, and their actual constraints. That is not just better output. That is a different category of work.
And that gap is going to widen fast.
The role is not disappearing. But it is splitting. I see it this way:
👨🏻💼 On one side, the PMs who carry context in their heads, re-explain their product every session, and produce work that sounds like everyone else’s work.
👩🏼💻 On the other side, the PMs who architect systems that remember for them, so they can spend every working hour on the decisions only they can make.
The prompts in this post are the beginning of a different way of working. One where your AI is not a tool you pick up and put down. It is a system that is already running when you sit down.
The vault is the first brick. What you build on top of it is up to you.
What is the first spec you are dropping into your vault today?
All prompts referenced in this post are available at promptledproduct.com/vault. Free prompts are accessible without a subscription. Paid prompts require a Prompt-Led Product paid subscription.
I Break AI Tools So You Don’t Have To
If you’re building a product and want a Senior PM to find the exact point where your users give up, I’m taking on new clients. I’ll check your architecture and test your logic before your users do. You can book a Product Audit here.
Stop chasing AI certificates. You don’t need another badge to prove you know your stuff. You need to build the thing!
At Cozora, we focus on building the skills you actually use at work. These are the high-paying skills that matter right now. It’s about execution, not theory.
I want to make sure you have the right tools to build, so here is the deal:
Paid subscribers: 50% off (monthly or annual)
Free subscribers: 10% off
Stop collecting credentials and start building systems. 🚀